Package 'coin'

General Information on the coin Package

Description

The coin package provides an implementation of a general framework for conditional inference procedures commonly known as permutation tests. The framework was developed by Strasser and Weber (1999) and is based on a multivariate linear statistic and its conditional expectation, covariance and limiting distribution. These results are utilized to construct tests of independence between two sets of variables.

The package does not only provide a flexible implementation of the abstract framework, but also provides a large set of convenience functions implementing well-known as well as lesser-known classical and non-classical test procedures within the framework. Many of the tests presented in prominent text books, such as Hollander and Wolfe (1999) or Agresti (2002), are immediately available or can be implemented without much effort. Examples include linear rank statistics for the two- and $K$-sample location and scale problem against ordered and unordered alternatives including post-hoc tests for arbitrary contrasts, tests of independence for contingency tables, two- and $K$-sample tests for censored data, tests of independence between two continuous variables as well as tests of marginal homogeneity and symmetry. Approximations of the exact null distribution via the limiting distribution or conditional Monte Carlo resampling are available for every test procedure, while the exact null distribution is currently available for univariate two-sample problems only.

The salient parts of the Strasser-Weber framework are elucidated by Hothorn et al. (2006) and a thorough description of the software implementation is given by Hothorn et al. (2008).

Author(s)

This package is authored by
Torsten Hothorn <[email protected]>,
Kurt Hornik <[email protected]>,
Mark A. van de Wiel <[email protected]>,
Henric Winell <[email protected]> and
Achim Zeileis <[email protected]>.

References

Agresti, A. (2002). Categorical Data Analysis, Second Edition. Hoboken, New Jersey: John Wiley & Sons.

Hollander, M. and Wolfe, D. A. (1999). Nonparametric Statistical Methods, Second Edition. New York: John Wiley & Sons.

Hothorn, T., Hornik, K., van de Wiel, M. A. and Zeileis, A. (2006). A Lego system for conditional inference. The American Statistician 60(3), 257–263. doi:10.1198/000313006X118430

Hothorn, T., Hornik, K., van de Wiel, M. A. and Zeileis, A. (2008). Implementing a class of permutation tests: The coin package. Journal of Statistical Software 28(8), 1–23. doi:10.18637/jss.v028.i08

Strasser, H. and Weber, C. (1999). On the asymptotic theory of permutation statistics. Mathematical Methods of Statistics 8(2), 220–250.

Examples

## Not run: 
## Generate doxygen documentation if you are interested in the internals:
## Download source package into a temporary directory
tmpdir <- tempdir()
tgz <- download.packages("coin", destdir = tmpdir, type = "source")[2]
## Extract contents
untar(tgz, exdir = tmpdir)
## Run doxygen (assuming it is installed)
wd <- setwd(file.path(tmpdir, "coin"))
system("doxygen inst/doxygen.cfg")
setwd(wd)
## Have fun!
browseURL(file.path(tmpdir, "coin", "inst",
                    "documentation", "html", "index.html"))
## End(Not run)

---

Genetic Components of Alcoholism

Description

Levels of expressed alpha synuclein mRNA in three groups of allele lengths of NACP-REP1.

Usage

alpha

Format

A data frame with 97 observations on 2 variables.

alength

allele length, a factor with levels "short", "intermediate" and "long".

elevel

expression levels of alpha synuclein mRNA.

Details

Various studies have linked alcohol dependence phenotypes to chromosome 4. One candidate gene is NACP (non-amyloid component of plaques), coding for alpha synuclein. Bönsch et al. (2005) found longer alleles of NACP-REP1 in alcohol-dependent patients compared with healthy controls and reported that the allele lengths show some association with levels of expressed alpha synuclein mRNA.

Source

Bönsch, D., Lederer, T., Reulbach, U., Hothorn, T., Kornhuber, J. and Bleich, S. (2005). Joint analysis of the NACP-REP1 marker within the alpha synuclein gene concludes association with alcohol dependence. Human Molecular Genetics 14(7), 967–971. doi:10.1093/hmg/ddi090

References

Hothorn, T., Hornik, K., van de Wiel, M. A. and Zeileis, A. (2006). A Lego system for conditional inference. The American Statistician 60(3), 257–263. doi:10.1198/000313006X118430

Winell, H. and Lindbäck, J. (2018). A general score-independent test for order-restricted inference. Statistics in Medicine 37(21), 3078–3090. doi:10.1002/sim.7690

Examples

## Boxplots
boxplot(elevel ~ alength, data = alpha)

## Asymptotic Kruskal-Wallis test
kruskal_test(elevel ~ alength, data = alpha)

## Asymptotic Kruskal-Wallis test using midpoint scores
kruskal_test(elevel ~ alength, data = alpha,
             scores = list(alength = c(2, 7, 11)))

## Asymptotic score-independent test
## Winell and Lindbaeck (2018)
(it <- independence_test(elevel ~ alength, data = alpha,
                         ytrafo = function(data)
                             trafo(data, numeric_trafo = rank_trafo), 
                         xtrafo = function(data)
                             trafo(data, factor_trafo = function(x)
                                 zheng_trafo(as.ordered(x)))))

## Extract the "best" set of scores
ss <- statistic(it, type = "standardized")
idx <- which(abs(ss) == max(abs(ss)), arr.ind = TRUE)
ss[idx[1], idx[2], drop = FALSE]

---

Smoking and Alzheimer's Disease

Description

A case-control study of smoking and Alzheimer's disease.

Usage

alzheimer

Format

A data frame with 538 observations on 3 variables.

smoking

a factor with levels "None", "<10", "10-20" and ">20" (cigarettes per day).

disease

a factor with levels "Alzheimer", "Other dementias" and "Other diagnoses".

gender

a factor with levels "Female" and "Male".

Details

Subjects with Alzheimer's disease are compared to two different control groups with respect to smoking history. The data are given in Salib and Hillier (1997, Tab. 4).

Source

Salib, E. and Hillier, V. (1997). A case-control study of smoking and Alzheimer's disease. International Journal of Geriatric Psychiatry 12(3), 295–300. doi:10.1002/(SICI)1099-1166(199703)12:3<295::AID-GPS476>3.0.CO;2-3

References

Hothorn, T., Hornik, K., van de Wiel, M. A. and Zeileis, A. (2006). A Lego system for conditional inference. The American Statistician 60(3), 257–263. doi:10.1198/000313006X118430

Examples

## Spineplots
op <- par(no.readonly = TRUE) # save current settings
layout(matrix(1:2, ncol = 2))
spineplot(disease ~ smoking, data = alzheimer,
          subset = gender == "Male", main = "Male")
spineplot(disease ~ smoking, data = alzheimer,
          subset = gender == "Female", main = "Female")
par(op) # reset

## Asymptotic Cochran-Mantel-Haenszel test
cmh_test(disease ~ smoking | gender, data = alzheimer)

---

Toxicological Study on Female Wistar Rats

Description

Measurements of the liver enzyme aspartate aminotransferase (ASAT) for a new compound and a control group of 34 female Wistar rats.

Usage

asat

Format

A data frame with 34 observations on 2 variables.

asat

ASAT values.

group

a factor with levels "Compound" and "Control".

Details

The aim of this toxicological study is the proof of safety for the new compound. The data were originally given in Hothorn (1992) and later reproduced by Hauschke, Kieser and Hothorn (1999).

Source

Hauschke, D., Kieser, M. and Hothorn, L. A. (1999). Proof of safety in toxicology based on the ratio of two means for normally distributed data. Biometrical Journal 41(3), 295–304. doi:10.1002/(SICI)1521-4036(199906)41:3<295::AID-BIMJ295>3.0.CO;2-2

Hothorn, L. A. (1992). Biometrische analyse toxikologischer untersuchungen. In J. Adam (Ed.), Statistisches Know-How in der Medizinischen Forschung, pp. 475–590. Berlin: Ullstein Mosby.

References

Pflüger, R. and Hothorn, T. (2002). Assessing equivalence tests with respect to their expected $p$-value. Biometrical Journal 44(8), 1015–1027. doi:10.1002/bimj.200290001

Examples

## Proof-of-safety based on ratio of medians (Pflueger and Hothorn, 2002)
## One-sided exact Wilcoxon-Mann-Whitney test
wt <- wilcox_test(I(log(asat)) ~ group, data = asat,
                  distribution = "exact", alternative = "less",
                  conf.int = TRUE)

## One-sided confidence set
## Note: Safety cannot be concluded since the effect of the compound
##       exceeds 20 % of the control median
exp(confint(wt)$conf.int)

---

Tests of Independence in Two- or Three-Way Contingency Tables

Description

Testing the independence of two nominal or ordered factors.

Usage

## S3 method for class 'formula'
chisq_test(formula, data, subset = NULL, weights = NULL, ...)
## S3 method for class 'table'
chisq_test(object, ...)
## S3 method for class 'IndependenceProblem'
chisq_test(object, ...)

## S3 method for class 'formula'
cmh_test(formula, data, subset = NULL, weights = NULL, ...)
## S3 method for class 'table'
cmh_test(object, ...)
## S3 method for class 'IndependenceProblem'
cmh_test(object, ...)

## S3 method for class 'formula'
lbl_test(formula, data, subset = NULL, weights = NULL, ...)
## S3 method for class 'table'
lbl_test(object, ...)
## S3 method for class 'IndependenceProblem'
lbl_test(object, ...)

Arguments

formula	a formula of the form y ~ x | block where y and x are factors and block is an optional factor for stratification.
data	an optional data frame containing the variables in the model formula.
subset	an optional vector specifying a subset of observations to be used. Defaults to NULL.
weights	an optional formula of the form ~ w defining integer valued case weights for each observation. Defaults to NULL, implying equal weight for all observations.
object	an object inheriting from classes "table" or "IndependenceProblem".
...	further arguments to be passed to independence_test().

Details

chisq_test(), cmh_test() and lbl_test() provide the Pearson chi-squared test, the generalized Cochran-Mantel-Haenszel test and the linear-by-linear association test. A general description of these methods is given by Agresti (2002).

The null hypothesis of independence, or conditional independence given block, between y and x is tested.

If y and/or x are ordered factors, the default scores, 1:nlevels(y) and 1:nlevels(x), respectively, can be altered using the scores argument (see independence_test()); this argument can also be used to coerce nominal factors to class "ordered". (lbl_test() coerces to class "ordered" under any circumstances.) If both y and x are ordered factors, a linear-by-linear association test is computed and the direction of the alternative hypothesis can be specified using the alternative argument. For the Pearson chi-squared test, this extension was given by Yates (1948) who also discussed the situation when either the response or the covariate is an ordered factor; see also Cochran (1954) and Armitage (1955) for the particular case when y is a binary factor and x is ordered. The Mantel-Haenszel statistic (Mantel and Haenszel, 1959) was similarly extended by Mantel (1963) and Landis, Heyman and Koch (1978).

The conditional null distribution of the test statistic is used to obtain $p$-values and an asymptotic approximation of the exact distribution is used by default (distribution = "asymptotic"). Alternatively, the distribution can be approximated via Monte Carlo resampling or computed exactly for univariate two-sample problems by setting distribution to "approximate" or "exact", respectively. See asymptotic(), approximate() and exact() for details.

Value

An object inheriting from class "IndependenceTest".

Note

The exact versions of the Pearson chi-squared test and the generalized Cochran-Mantel-Haenszel test do not necessarily result in the same $p$-value as Fisher's exact test (Davis, 1986).

References

Agresti, A. (2002). Categorical Data Analysis, Second Edition. Hoboken, New Jersey: John Wiley & Sons.

Armitage, P. (1955). Tests for linear trends in proportions and frequencies. Biometrics 11(3), 375–386. doi:10.2307/3001775

Cochran, W.G. (1954). Some methods for strengthening the common $\chi^2$ tests. Biometrics 10(4), 417–451. doi:10.2307/3001616

Davis, L. J. (1986). Exact tests for $2 \times 2$ contingency tables. The American Statistician 40(2), 139–141. doi:10.1080/00031305.1986.10475377

Landis, J. R., Heyman, E. R. and Koch, G. G. (1978). Average partial association in three-way contingency tables: a review and discussion of alternative tests. International Statistical Review 46(3), 237–254. doi:10.2307/1402373

Mantel, N. and Haenszel, W. (1959). Statistical aspects of the analysis of data from retrospective studies of disease. Journal of the National Cancer Institute 22(4), 719–748. doi:10.1093/jnci/22.4.719

Mantel, N. (1963). Chi-square tests with one degree of freedom: extensions of the Mantel-Haenszel procedure. Journal of the American Statistical Association 58(303), 690–700. doi:10.1080/01621459.1963.10500879

Yates, F. (1948). The analysis of contingency tables with groupings based on quantitative characters. Biometrika 35(1/2), 176–181. doi:10.1093/biomet/35.1-2.176

Examples

## Example data
## Davis (1986, p. 140)
davis <- matrix(
    c(3,  6,
      2, 19),
    nrow = 2, byrow = TRUE
)
davis <- as.table(davis)

## Asymptotic Pearson chi-squared test
chisq_test(davis)
chisq.test(davis, correct = FALSE) # same as above

## Approximative (Monte Carlo) Pearson chi-squared test
ct <- chisq_test(davis,
                 distribution = approximate(nresample = 10000))
pvalue(ct)             # standard p-value
midpvalue(ct)          # mid-p-value
pvalue_interval(ct)    # p-value interval
size(ct, alpha = 0.05) # test size at alpha = 0.05 using the p-value

## Exact Pearson chi-squared test (Davis, 1986)
## Note: disagrees with Fisher's exact test
ct <- chisq_test(davis,
                 distribution = "exact")
pvalue(ct)             # standard p-value
midpvalue(ct)          # mid-p-value
pvalue_interval(ct)    # p-value interval
size(ct, alpha = 0.05) # test size at alpha = 0.05 using the p-value
fisher.test(davis)


## Laryngeal cancer data
## Agresti (2002, p. 107, Tab. 3.13)
cancer <- matrix(
    c(21, 2,
      15, 3),
    nrow = 2, byrow = TRUE,
    dimnames = list(
        "Treatment" = c("Surgery", "Radiation"),
           "Cancer" = c("Controlled", "Not Controlled")
    )
)
cancer <- as.table(cancer)

## Exact Pearson chi-squared test (Agresti, 2002, p. 108, Tab. 3.14)
## Note: agrees with Fishers's exact test
(ct <- chisq_test(cancer,
                  distribution = "exact"))
midpvalue(ct)          # mid-p-value
pvalue_interval(ct)    # p-value interval
size(ct, alpha = 0.05) # test size at alpha = 0.05 using the p-value
fisher.test(cancer)


## Homework conditions and teacher's rating
## Yates (1948, Tab. 1)
yates <- matrix(
    c(141, 67, 114, 79, 39,
      131, 66, 143, 72, 35,
       36, 14,  38, 28, 16),
    byrow = TRUE, ncol = 5,
    dimnames = list(
           "Rating" = c("A", "B", "C"),
        "Condition" = c("A", "B", "C", "D", "E")
    )
)
yates <- as.table(yates)

## Asymptotic Pearson chi-squared test (Yates, 1948, p. 176)
chisq_test(yates)

## Asymptotic Pearson-Yates chi-squared test (Yates, 1948, pp. 180-181)
## Note: 'Rating' and 'Condition' as ordinal
(ct <- chisq_test(yates,
                  alternative = "less",
                  scores = list("Rating" = c(-1, 0, 1),
                                "Condition" = c(2, 1, 0, -1, -2))))
statistic(ct)^2 # chi^2 = 2.332

## Asymptotic Pearson-Yates chi-squared test (Yates, 1948, p. 181)
## Note: 'Rating' as ordinal
chisq_test(yates,
           scores = list("Rating" = c(-1, 0, 1))) # Q = 3.825


## Change in clinical condition and degree of infiltration
## Cochran (1954, Tab. 6)
cochran <- matrix(
    c(11,  7,
      27, 15,
      42, 16,
      53, 13,
      11,  1),
    byrow = TRUE, ncol = 2,
    dimnames = list(
              "Change" = c("Marked", "Moderate", "Slight",
                           "Stationary", "Worse"),
        "Infiltration" = c("0-7", "8-15")
    )
)
cochran <- as.table(cochran)

## Asymptotic Pearson chi-squared test (Cochran, 1954, p. 435)
chisq_test(cochran) # X^2 = 6.88

## Asymptotic Cochran-Armitage test (Cochran, 1954, p. 436)
## Note: 'Change' as ordinal
(ct <- chisq_test(cochran,
                  scores = list("Change" = c(3, 2, 1, 0, -1))))
statistic(ct)^2 # X^2 = 6.66


## Change in size of ulcer crater for two treatment groups
## Armitage (1955, Tab. 2)
armitage <- matrix(
    c( 6, 4, 10, 12,
      11, 8,  8,  5),
    byrow = TRUE, ncol = 4,
    dimnames = list(
        "Treatment" = c("A", "B"),
           "Crater" = c("Larger", "< 2/3 healed",
                        ">= 2/3 healed", "Healed")
    )
)
armitage <- as.table(armitage)

## Approximative (Monte Carlo) Pearson chi-squared test (Armitage, 1955, p. 379)
chisq_test(armitage,
           distribution = approximate(nresample = 10000)) # chi^2 = 5.91

## Approximative (Monte Carlo) Cochran-Armitage test (Armitage, 1955, p. 379)
(ct <- chisq_test(armitage,
                  distribution = approximate(nresample = 10000),
                  scores = list("Crater" = c(-1.5, -0.5, 0.5, 1.5))))
statistic(ct)^2 # chi_0^2 = 5.26


## Relationship between job satisfaction and income stratified by gender
## Agresti (2002, p. 288, Tab. 7.8)

## Asymptotic generalized Cochran-Mantel-Haenszel test (Agresti, p. 297)
(ct <- cmh_test(jobsatisfaction)) # CMH = 10.2001

## The standardized linear statistic
statistic(ct, type = "standardized")

## The standardized linear statistic for each block
statistic(ct, type = "standardized", partial = TRUE)

## Asymptotic generalized Cochran-Mantel-Haenszel test (Agresti, p. 297)
## Note: 'Job.Satisfaction' as ordinal
cmh_test(jobsatisfaction,
         scores = list("Job.Satisfaction" = c(1, 3, 4, 5))) # L^2 = 9.0342

## Asymptotic linear-by-linear association test (Agresti, p. 297)
## Note: 'Job.Satisfaction' and 'Income' as ordinal
(lt <- lbl_test(jobsatisfaction,
                scores = list("Job.Satisfaction" = c(1, 3, 4, 5),
                              "Income" = c(3, 10, 20, 35))))
statistic(lt)^2 # M^2 = 6.1563

## The standardized linear statistic
statistic(lt, type = "standardized")

## The standardized linear statistic for each block
statistic(lt, type = "standardized", partial = TRUE)

---

Correlation Tests

Description

Testing the independence of two numeric variables.

Usage

## S3 method for class 'formula'
spearman_test(formula, data, subset = NULL, weights = NULL, ...)
## S3 method for class 'IndependenceProblem'
spearman_test(object, distribution = c("asymptotic", "approximate", "none"), ...)

## S3 method for class 'formula'
fisyat_test(formula, data, subset = NULL, weights = NULL, ...)
## S3 method for class 'IndependenceProblem'
fisyat_test(object, distribution = c("asymptotic", "approximate", "none"),
            ties.method = c("mid-ranks", "average-scores"), ...)

## S3 method for class 'formula'
quadrant_test(formula, data, subset = NULL, weights = NULL, ...)
## S3 method for class 'IndependenceProblem'
quadrant_test(object, distribution = c("asymptotic", "approximate", "none"),
              mid.score = c("0", "0.5", "1"), ...)

## S3 method for class 'formula'
koziol_test(formula, data, subset = NULL, weights = NULL, ...)
## S3 method for class 'IndependenceProblem'
koziol_test(object, distribution = c("asymptotic", "approximate", "none"),
            ties.method = c("mid-ranks", "average-scores"), ...)

Arguments

formula	a formula of the form y ~ x | block where y and x are numeric variables and block is an optional factor for stratification.
data	an optional data frame containing the variables in the model formula.
subset	an optional vector specifying a subset of observations to be used. Defaults to NULL.
weights	an optional formula of the form ~ w defining integer valued case weights for each observation. Defaults to NULL, implying equal weight for all observations.
object	an object inheriting from class "IndependenceProblem".
distribution	a character, the conditional null distribution of the test statistic can be approximated by its asymptotic distribution ("asymptotic", default) or via Monte Carlo resampling ("approximate"). Alternatively, the functions asymptotic or approximate can be used. Computation of the null distribution can be suppressed by specifying "none".
ties.method	a character, the method used to handle ties: the score generating function either uses mid-ranks ("mid-ranks", default) or averages the scores of randomly broken ties ("average-scores").
mid.score	a character, the score assigned to observations exactly equal to the median: either 0 ("0", default), 0.5 ("0.5") or 1 ("1"); see median_test().
...	further arguments to be passed to independence_test().

Details

spearman_test(), fisyat_test(), quadrant_test() and koziol_test() provide the Spearman correlation test, the Fisher-Yates correlation test using van der Waerden scores, the quadrant test and the Koziol-Nemec test. A general description of these methods is given by Hájek, Šidák and Sen (1999, Sec. 4.6). The Koziol-Nemec test was suggested by Koziol and Nemec (1979). For the adjustment of scores for tied values see Hájek, Šidák and Sen (1999, pp. 133–135).

The null hypothesis of independence, or conditional independence given block, between y and x is tested.

The conditional null distribution of the test statistic is used to obtain $p$-values and an asymptotic approximation of the exact distribution is used by default (distribution = "asymptotic"). Alternatively, the distribution can be approximated via Monte Carlo resampling by setting distribution to "approximate". See asymptotic() and approximate() for details.

Value

An object inheriting from class "IndependenceTest".

References

Hájek, J., Šidák, Z. and Sen, P. K. (1999). Theory of Rank Tests, Second Edition. San Diego: Academic Press.

Koziol, J. A. and Nemec, A. F. (1979). On a Cramér-von Mises type statistic for testing bivariate independence. The Canadian Journal of Statistics 7(1), 43–52. doi:10.2307/3315014

Examples

## Asymptotic Spearman test
spearman_test(CONT ~ INTG, data = USJudgeRatings)

## Asymptotic Fisher-Yates test
fisyat_test(CONT ~ INTG, data = USJudgeRatings)

## Asymptotic quadrant test
quadrant_test(CONT ~ INTG, data = USJudgeRatings)

## Asymptotic Koziol-Nemec test
koziol_test(CONT ~ INTG, data = USJudgeRatings)

---

Coarse Woody Debris

Description

Carbon flux on six pieces of wood.

Usage

CWD

Format

A data frame with 13 observations on 8 variables.

sample2

carbon flux measurement for 2nd piece of wood.

sample3

carbon flux measurement for 3rd piece of wood.

sample4

carbon flux measurement for 4th piece of wood.

sample6

carbon flux measurement for 6th piece of wood.

sample7

carbon flux measurement for 7th piece of wood.

sample8

carbon flux measurement for 8th piece of wood.

trend

measurement day (in days from beginning).

time

date of measurement.

Details

Coarse woody debris (CWD, dead wood greater than 10 cm in diameter) is a large stock of carbon in tropical forests, yet the flux of carbon out of this pool, via respiration, is poorly resolved (Chambers, Schimel and Nobre, 2001). The heterotrophic process involved in CWD respiration should respond to reductions in moisture availability, which occurs during dry season (Chambers, Schimel and Nobre, 2001).

CWD respiration measurements were taken in a tropical forest in west French Guiana, which experiences extreme contrasts in wet and dry season (Bonal et al., 2008). An infrared gas analyzer and a clear chamber sealed to the wood surface were used to measure the flux of carbon out of the wood (Stahl et al., 2011). Measurements were repeated 13 times, from July to November 2011, on six pieces of wood during the transition into and out of the dry season. The aim is to assess if there were shifts in the CWD respiration of any of the pieces in response to the transition into (early August) and out of (late October) the dry season.

Zeileis and Hothorn (2013) investigated the six-variate series of CO$_2$ reflux, aiming to find out whether the reflux had changed over the sampling period in at least one of the six wood pieces.

Source

The coarse woody debris respiration data were kindly provided by Lucy Rowland (School of GeoSciences, University of Edinburgh).

References

Bonal, D., Bosc, A., Ponton, S., Goret, J.-Y., Burban, B., Gross, P., Bonnefond, J.-M., Elbers, J., Longdoz, B., Epron, D., Guehl, J.-M. and Granier, A. (2008). Impact of severe dry season on net ecosystem exchange in the Neotropical rainforest of French Guiana. Global Change Biology 14(8), 1917–1933. doi:10.1111/j.1365-2486.2008.01610.x

Chambers, J. Q., Schimel, J. P. and Nobre, A. D. (2001). Respiration from coarse wood litter in central Amazon forests. Biogeochemistry 52(2), 115–131. doi:10.1023/A:1006473530673

Stahl, C., Burban, B., Goret, J.-Y. and Bonal, D. (2011). Seasonal variations in stem CO$\mathrm{_2}$ efflux in the Neotropical rainforest of French Guiana. Annals of Forest Science 68(4), 771–782. doi:10.1007/s13595-011-0074-2

Zeileis, A. and Hothorn, T. (2013). A toolbox of permutation tests for structural change. Statistical Papers 54(4), 931–954. doi:10.1007/s00362-013-0503-4

Examples

## Zeileis and Hothorn (2013, pp. 942-944)
## Approximative (Monte Carlo) maximally selected statistics
CWD[1:6] <- 100 * CWD[1:6] # scaling (to avoid harmless warning)
mt <- maxstat_test(sample2 + sample3 + sample4 +
                   sample6 + sample7 + sample8 ~ trend, data = CWD,
                   distribution = approximate(nresample = 100000))

## Absolute maximum of standardized statistics (t = 3.08)
statistic(mt)

## 5% critical value (t_0.05 = 2.86)
(c <- qperm(mt, 0.95))

## Only 'sample8' exceeds the 5% critical value
sts <- statistic(mt, type = "standardized")
idx <- which(sts > c, arr.ind = TRUE)
sts[unique(idx[, 1]), unique(idx[, 2]), drop = FALSE]

---

Extraction of the Expectation, Variance and Covariance of the Linear Statistic

Description

Methods for extraction of the expectation, variance and covariance of the linear statistic.

Usage

## S4 method for signature 'IndependenceLinearStatistic'
expectation(object, partial = FALSE, ...)
## S4 method for signature 'IndependenceTest'
expectation(object, partial = FALSE, ...)

## S4 method for signature 'Variance'
variance(object, ...)
## S4 method for signature 'CovarianceMatrix'
variance(object, ...)
## S4 method for signature 'IndependenceLinearStatistic'
variance(object, partial = FALSE, ...)
## S4 method for signature 'IndependenceTest'
variance(object, partial = FALSE, ...)

## S4 method for signature 'CovarianceMatrix'
covariance(object, ...)
## S4 method for signature 'IndependenceLinearStatistic'
covariance(object, invert = FALSE, partial = FALSE, ...)
## S4 method for signature 'QuadTypeIndependenceTestStatistic'
covariance(object, invert = FALSE, partial = FALSE, ...)
## S4 method for signature 'IndependenceTest'
covariance(object, invert = FALSE, partial = FALSE, ...)

Arguments

object	an object from which the expectation, variance or covariance of the linear statistic can be extracted.
partial	a logical indicating that the partial result for each block should be extracted. Defaults to FALSE.
invert	a logical indicating that the Moore-Penrose inverse of the covariance should be extracted. Defaults to FALSE.
...	further arguments (currently ignored).

Details

The methods expectation, variance and covariance extract the expectation, variance and covariance, respectively, of the linear statistic.

For tests of conditional independence within blocks, the partial result for each block is obtained by setting partial = TRUE.

Value

The expectation, variance or covariance of the linear statistic extracted from object. A matrix or array.

Examples

## Example data
dta <- data.frame(
    y = gl(3, 2),
    x = sample(gl(3, 2))
)

## Asymptotic Cochran-Mantel-Haenszel Test
ct <- cmh_test(y ~ x, data = dta)

## The linear statistic, i.e., the contingency table...
(T <- statistic(ct, type = "linear"))

## ...and its expectation...
(mu <- expectation(ct))

## ...and variance...
(sigma <- variance(ct))

## ...and covariance...
(Sigma <- covariance(ct))

## ...and its inverse
(SigmaPlus <- covariance(ct, invert = TRUE))

## The standardized contingency table...
(T - mu) / sqrt(sigma)

## ...is identical to the standardized linear statistic
statistic(ct, type = "standardized")

## The quadratic form...
U <- as.vector(T - mu)
U %*% SigmaPlus %*% U

## ...is identical to the test statistic
statistic(ct, type = "test")

---

Malignant Glioma Pilot Study

Description

A non-randomized pilot study on malignant glioma patients with pretargeted adjuvant radioimmunotherapy using yttrium-90-biotin.

Usage

glioma

Format

A data frame with 37 observations on 7 variables.

no.

patient number.

age

patient age (years).

sex

a factor with levels "F" (Female) and "M" (Male).

histology

a factor with levels "GBM" (grade IV) and "Grade3" (grade III).

group

a factor with levels "Control" and "RIT".

event

status indicator for time: FALSE for right-censored observations and TRUE otherwise.

time

survival time (months).

Details

The primary endpoint of this small pilot study is survival. Since the survival times are tied, the classical asymptotic logrank test may be inadequate in this setup. Therefore, a permutation test using Monte Carlo resampling was computed in the original paper. The data are taken from Tables 1 and 2 of Grana et al. (2002).

Source

Grana, C., Chinol, M., Robertson, C., Mazzetta, C., Bartolomei, M., De Cicco, C., Fiorenza, M., Gatti, M., Caliceti, P. and Paganelli, G. (2002). Pretargeted adjuvant radioimmunotherapy with Yttrium-90-biotin in malignant glioma patients: A pilot study. British Journal of Cancer 86(2), 207–212. doi:10.1038/sj.bjc.6600047

Examples

## Grade III glioma
g3 <- subset(glioma, histology == "Grade3")

## Plot Kaplan-Meier estimates
op <- par(no.readonly = TRUE) # save current settings
layout(matrix(1:2, ncol = 2))
plot(survfit(Surv(time, event) ~ group, data = g3),
     main = "Grade III Glioma", lty = 2:1,
     ylab = "Probability", xlab = "Survival Time in Month",
     xlim = c(-2, 72))
legend("bottomleft", lty = 2:1, c("Control", "Treated"), bty = "n")

## Exact logrank test
logrank_test(Surv(time, event) ~ group, data = g3,
             distribution = "exact")


## Grade IV glioma
gbm <- subset(glioma, histology == "GBM")

## Plot Kaplan-Meier estimates
plot(survfit(Surv(time, event) ~ group, data = gbm),
     main = "Grade IV Glioma", lty = 2:1,
     ylab = "Probability", xlab = "Survival Time in Month",
     xlim = c(-2, 72))
legend("topright", lty = 2:1, c("Control", "Treated"), bty = "n")
par(op) # reset

## Exact logrank test
logrank_test(Surv(time, event) ~ group, data = gbm,
             distribution = "exact")


## Stratified approximative (Monte Carlo) logrank test
logrank_test(Surv(time, event) ~ group | histology, data = glioma,
             distribution = approximate(nresample = 10000))

---

Gastrointestinal Tumor Study Group

Description

A randomized clinical trial in gastric cancer.

Usage

GTSG

Format

A data frame with 90 observations on 3 variables.

time

survival time (days).

event

status indicator for time: 0 for right-censored observations and 1 otherwise.

group

a factor with levels "Chemotherapy+Radiation" and "Chemotherapy".

Details

A clinical trial comparing chemotherapy alone versus a combination of chemotherapy and radiation therapy in the treatment of locally advanced, nonresectable gastric carcinoma.

Note

There is substantial separation between the estimated survival distributions at 8 to 10 months, but by month 26 the distributions intersect.

Source

Stablein, D. M., Carter, W. H., Jr. and Novak, J. W. (1981). Analysis of survival data with nonproportional hazard functions. Controlled Clinical Trials 2(2), 149–159. doi:10.1016/0197-2456(81)90005-2

References

Moreau, T., Maccario, J., Lellouch, J. and Huber, C. (1992). Weighted log rank statistics for comparing two distributions. Biometrika 79(1), 195–198. doi:10.1093/biomet/79.1.195

Shen, W. and Le, C. T. (2000). Linear rank tests for censored survival data. Communications in Statistics – Simulation and Computation 29(1), 21–36. doi:10.1080/03610910008813599

Tubert-Bitter, P., Kramar, A., Chalé, J. J. and Moureau, T. (1994). Linear rank tests for comparing survival in two groups with crossing hazards. Computational Statistics & Data Analysis 18(5), 547–559. doi:10.1016/0167-9473(94)90084-1

Examples

## Plot Kaplan-Meier estimates
plot(survfit(Surv(time / (365.25 / 12), event) ~ group, data = GTSG),
     lty = 1:2, ylab = "% Survival", xlab = "Survival Time in Months")
legend("topright", lty = 1:2,
       c("Chemotherapy+Radiation", "Chemotherapy"), bty = "n")

## Asymptotic logrank test
logrank_test(Surv(time, event) ~ group, data = GTSG)

## Asymptotic Prentice test
logrank_test(Surv(time, event) ~ group, data = GTSG, type = "Prentice")

## Asymptotic test against Weibull-type alternatives (Moreau et al., 1992)
moreau_weight <- function(time, n.risk, n.event)
    1 + log(-log(cumprod(n.risk / (n.risk + n.event))))

independence_test(Surv(time, event) ~ group, data = GTSG,
                  ytrafo = function(data)
                      trafo(data, surv_trafo = function(y)
                          logrank_trafo(y, weight = moreau_weight)))

## Asymptotic test against crossing-curve alternatives (Shen and Le, 2000)
shen_trafo <- function(x)
    ansari_trafo(logrank_trafo(x, type = "Prentice"))

independence_test(Surv(time, event) ~ group, data = GTSG,
                  ytrafo = function(data)
                      trafo(data, surv_trafo = shen_trafo))

---

Left Ventricular Ejection Fraction

Description

Left ventricular ejection fraction in patients with malignant ventricular tachyarrhythmias including recurrence-free month and censoring.

Usage

hohnloser

Format

A data frame with 94 observations on 3 variables.

EF

ejection fraction (%).

time

recurrence-free month.

event

status indicator for time: 0 for right-censored observations and 1 otherwise.

Details

The data was used by Lausen and Schumacher (1992) to illustrate the use of maximally selected statistics.

Source

Hohnloser, S. H., Raeder, E. A., Podrid, P. J., Graboys, T. B. and Lown, B. (1987). Predictors of antiarrhythmic drug efficacy in patients with malignant ventricular tachyarrhythmias. American Heart Journal 114(1 Pt 1), 1–7. doi:10.1016/0002-8703(87)90299-7

References

Lausen, B. and Schumacher, M. (1992). Maximally selected rank statistics. Biometrics 48(1), 73–85. doi:10.2307/2532740

Examples

## Asymptotic maximally selected logrank statistics
maxstat_test(Surv(time, event) ~ EF, data = hohnloser)

---

Class "IndependenceLinearStatistic"

Description

Objects of class "IndependenceLinearStatistic" represent the linear statistic and the transformed and original data structures corresponding to an independence problem.

Objects from the Class

Objects can be created by calls of the form

     new("IndependenceLinearStatistic", object, ...)

where object is an object of class "IndependenceTestProblem".

Slots

linearstatistic:

Object of class "matrix". The linear statistic for each block.

expectation:

Object of class "matrix". The expectation of the linear statistic for each block.

covariance:

Object of class "matrix". The lower triangular elements of the covariance of the linear statistic for each block.

xtrans:

Object of class "matrix". The transformed x.

ytrans:

Object of class "matrix". The transformed y.

xtrafo:

Object of class "function". The regression function for x.

ytrafo:

Object of class "function". The influence function for y.

x:

Object of class "data.frame". The variables x.

y:

Object of class "data.frame". The variables y.

block:

Object of class "factor". The block structure.

weights:

Object of class "numeric". The case weights.

Extends

Class "IndependenceTestProblem", directly.
Class "IndependenceProblem", by class "IndependenceTestProblem", distance 2.

Known Subclasses

Class "IndependenceTestStatistic", directly.
Class "MaxTypeIndependenceTestStatistic", by class "IndependenceTestStatistic", distance 2.
Class "QuadTypeIndependenceTestStatistic", by class "IndependenceTestStatistic", distance 2.
Class "ScalarIndependenceTestStatistic", by class "IndependenceTestStatistic", distance 2.

Methods

covariance

signature(object = "IndependenceLinearStatistic"): See the documentation for covariance() for details.

expectation

signature(object = "IndependenceLinearStatistic"): See the documentation for expectation() for details.

initialize

signature(.Object = "IndependenceLinearStatistic"): See the documentation for initialize() (in package methods) for details.

statistic

signature(object = "IndependenceLinearStatistic"): See the documentation for statistic() for details.

variance

signature(object = "IndependenceLinearStatistic"): See the documentation for variance() for details.

---

Class "IndependenceProblem"

Description

Objects of class "IndependenceProblem" represent the data structure corresponding to an independence problem.

Objects from the Class

Objects can be created by calls of the form

     new("IndependenceProblem", x, y, block = NULL, weights = NULL, ...)

where x and y are data frames containing the variables $\mathbf{X}$ and $\mathbf{Y}$, respectively, block is an optional factor representing the block structure $b$ and weights is an optional integer vector corresponding to the case weights $w$.

Slots

x:

Object of class "data.frame". The variables x.

y:

Object of class "data.frame". The variables y.

block:

Object of class "factor". The block structure.

weights:

Object of class "numeric". The case weights.

Known Subclasses

Class "IndependenceTestProblem", directly.
Class "SymmetryProblem", directly.
Class "IndependenceLinearStatistic", by class "IndependenceTestProblem", distance 2.
Class "IndependenceTestStatistic", by class "IndependenceTestProblem", distance 3.
Class "MaxTypeIndependenceTestStatistic", by class "IndependenceTestProblem", distance 4.
Class "QuadTypeIndependenceTestStatistic", by class "IndependenceTestProblem", distance 4.
Class "ScalarIndependenceTestStatistic", by class "IndependenceTestProblem", distance 4.

Methods

initialize

signature(.Object = "IndependenceProblem"): See the documentation for initialize() (in package methods) for details.

---

General Independence Test

Description

Testing the independence of two sets of variables measured on arbitrary scales.

Usage

## S3 method for class 'formula'
independence_test(formula, data, subset = NULL, weights = NULL, ...)
## S3 method for class 'table'
independence_test(object, ...)
## S3 method for class 'IndependenceProblem'
independence_test(object, teststat = c("maximum", "quadratic", "scalar"),
                  distribution = c("asymptotic", "approximate",
                                   "exact", "none"),
                  alternative = c("two.sided", "less", "greater"),
                  xtrafo = trafo, ytrafo = trafo, scores = NULL,
                  check = NULL, ...)

Arguments

formula	a formula of the form y1 + ... + yq ~ x1 + ... + xp | block where y1, ..., yq and x1, ..., xp are measured on arbitrary scales (nominal, ordinal or continuous with or without censoring) and block is an optional factor for stratification.
data	an optional data frame containing the variables in the model formula.
subset	an optional vector specifying a subset of observations to be used. Defaults to NULL.
weights	an optional formula of the form ~ w defining integer valued case weights for each observation. Defaults to NULL, implying equal weight for all observations.
object	an object inheriting from classes "table" or "IndependenceProblem".
teststat	a character, the type of test statistic to be applied: either a maximum statistic ("maximum", default), a quadratic form ("quadratic") or a standardized scalar test statistic ("scalar").
distribution	a character, the conditional null distribution of the test statistic can be approximated by its asymptotic distribution ("asymptotic", default) or via Monte Carlo resampling ("approximate"). Alternatively, the functions asymptotic or approximate can be used. For univariate two-sample problems, "exact" or use of the function exact computes the exact distribution. Computation of the null distribution can be suppressed by specifying "none". It is also possible to specify a function with one argument (an object inheriting from "IndependenceTestStatistic") that returns an object of class "NullDistribution".
alternative	a character, the alternative hypothesis: either "two.sided" (default), "greater" or "less".
xtrafo	a function of transformations to be applied to the variables x1, ..., xp supplied in formula; see ‘Details’. Defaults to trafo().
ytrafo	a function of transformations to be applied to the variables y1, ..., yq supplied in formula; see ‘Details’. Defaults to trafo().
scores	a named list of scores to be attached to ordered factors; see ‘Details’. Defaults to NULL, implying equally spaced scores.
check	a function to be applied to objects of class "IndependenceTest" in order to check for specific properties of the data. Defaults to NULL.
...	further arguments to be passed to or from other methods (currently ignored).

Details

independence_test() provides a general independence test for two sets of variables measured on arbitrary scales. This function is based on the general framework for conditional inference procedures proposed by Strasser and Weber (1999). The salient parts of the Strasser-Weber framework are elucidated by Hothorn et al. (2006) and a thorough description of the software implementation is given by Hothorn et al. (2008).

The null hypothesis of independence, or conditional independence given block, between y1, ..., yq and x1, ..., xp is tested.

A vector of case weights, e.g., observation counts, can be supplied through the weights argument and the type of test statistic is specified by the teststat argument. Influence and regression functions, i.e., transformations of y1, ..., yq and x1, ..., xp, are specified by the ytrafo and xtrafo arguments, respectively; see trafo() for the collection of transformation functions currently available. This allows for implementation of both novel and familiar test statistics, e.g., the Pearson $\chi^2$ test, the generalized Cochran-Mantel-Haenszel test, the Spearman correlation test, the Fisher-Pitman permutation test, the Wilcoxon-Mann-Whitney test, the Kruskal-Wallis test and the family of weighted logrank tests for censored data. Furthermore, multivariate extensions such as the multivariate Kruskal-Wallis test (Puri and Sen, 1966, 1971) can be implemented without much effort (see ‘Examples’).

If, say, y1 and/or x1 are ordered factors, the default scores, 1:nlevels(y1) and 1:nlevels(x1), respectively, can be altered using the scores argument; this argument can also be used to coerce nominal factors to class "ordered". For example, when y1 is an ordered factor with four levels and x1 is a nominal factor with three levels, scores = list(y1 = c(1, 3:5), x1 = c(1:2, 4)) supplies the scores to be used. For ordered alternatives the scores must be monotonic, but non-monotonic scores are also allowed for testing against, e.g., umbrella alternatives. The length of the score vector must be equal to the number of factor levels.

The conditional null distribution of the test statistic is used to obtain $p$-values and an asymptotic approximation of the exact distribution is used by default (distribution = "asymptotic"). Alternatively, the distribution can be approximated via Monte Carlo resampling or computed exactly for univariate two-sample problems by setting distribution to "approximate" or "exact", respectively. See asymptotic(), approximate() and exact() for details.

Value

An object inheriting from class "IndependenceTest".

Note

Starting with coin version 1.1-0, maximum statistics and quadratic forms can no longer be specified using teststat = "maxtype" and teststat = "quadtype", respectively (as was used in versions prior to 0.4-5).

References

Hothorn, T., Hornik, K., van de Wiel, M. A. and Zeileis, A. (2006). A Lego system for conditional inference. The American Statistician 60(3), 257–263. doi:10.1198/000313006X118430

Hothorn, T., Hornik, K., van de Wiel, M. A. and Zeileis, A. (2008). Implementing a class of permutation tests: The coin package. Journal of Statistical Software 28(8), 1–23. doi:10.18637/jss.v028.i08

Johnson, W. D., Mercante, D. E. and May, W. L. (1993). A computer package for the multivariate nonparametric rank test in completely randomized experimental designs. Computer Methods and Programs in Biomedicine 40(3), 217–225. doi:10.1016/0169-2607(93)90059-T

Puri, M. L. and Sen, P. K. (1966). On a class of multivariate multisample rank order tests. Sankhya A 28(4), 353–376.

Puri, M. L. and Sen, P. K. (1971). Nonparametric Methods in Multivariate Analysis. New York: John Wiley & Sons.

Strasser, H. and Weber, C. (1999). On the asymptotic theory of permutation statistics. Mathematical Methods of Statistics 8(2), 220–250.

Examples

## One-sided exact van der Waerden (normal scores) test...
independence_test(asat ~ group, data = asat,
                  ## exact null distribution
                  distribution = "exact",
                  ## one-sided test
                  alternative = "greater",
                  ## apply normal scores to asat$asat
                  ytrafo = function(data)
                      trafo(data, numeric_trafo = normal_trafo),
                  ## indicator matrix of 1st level of asat$group
                  xtrafo = function(data)
                      trafo(data, factor_trafo = function(x)
                          matrix(x == levels(x)[1], ncol = 1)))

## ...or more conveniently
normal_test(asat ~ group, data = asat,
            ## exact null distribution
            distribution = "exact",
            ## one-sided test
            alternative = "greater")


## Receptor binding assay of benzodiazepines
## Johnson, Mercante and May (1993, Tab. 1)
benzos <- data.frame(
      cerebellum = c( 3.41,  3.50,  2.85,  4.43,
                      4.04,  7.40,  5.63, 12.86,
                      6.03,  6.08,  5.75,  8.09,  7.56),
       brainstem = c( 3.46,  2.73,  2.22,  3.16,
                      2.59,  4.18,  3.10,  4.49,
                      6.78,  7.54,  5.29,  4.57,  5.39),
          cortex = c(10.52,  7.52,  4.57,  5.48,
                      7.16, 12.00,  9.36,  9.35,
                     11.54, 11.05,  9.92, 13.59, 13.21),
    hypothalamus = c(19.51, 10.00,  8.27, 10.26,
                     11.43, 19.13, 14.03, 15.59,
                     24.87, 14.16, 22.68, 19.93, 29.32),
        striatum = c( 6.98,  5.07,  3.57,  5.34,
                      4.57,  8.82,  5.76, 11.72,
                      6.98,  7.54,  7.66,  9.69,  8.09),
     hippocampus = c(20.31, 13.20,  8.58, 11.42,
                     13.79, 23.71, 18.35, 38.52,
                     21.56, 18.66, 19.24, 27.39, 26.55),
       treatment = factor(rep(c("Lorazepam", "Alprazolam", "Saline"),
                          c(4, 4, 5)))
)

## Approximative (Monte Carlo) multivariate Kruskal-Wallis test
## Johnson, Mercante and May (1993, Tab. 2)
independence_test(cerebellum + brainstem + cortex +
                  hypothalamus + striatum + hippocampus ~ treatment,
                  data = benzos,
                  teststat = "quadratic",
                  distribution = approximate(nresample = 10000),
                  ytrafo = function(data)
                      trafo(data, numeric_trafo = rank_trafo)) # Q = 16.129

---

Class "IndependenceTest" and Its Subclasses

Description

Objects of class "IndependenceTest" and its subclasses "MaxTypeIndependenceTest", "QuadTypeIndependenceTest", "ScalarIndependenceTest" and "ScalarIndependenceTestConfint" represent an independence test including its original and transformed data structure, linear statistic, test statistic and reference distribution.

Objects from the Class

Objects can be created by calls of the form

    new("IndependenceTest", ...),

    new("MaxTypeIndependenceTest", ...),

    new("QuadTypeIndependenceTest", ...),

    new("ScalarIndependenceTest", ...)

and

    new("ScalarIndependenceTestConfint", ...).

Slots

For objects of classes "IndependenceTest", "MaxTypeIndependenceTest", "QuadTypeIndependenceTest", "ScalarIndependenceTest" or "ScalarIndependenceTestConfint":

distribution:

Object of class "PValue". The reference distribution.

statistic:

Object of class "IndependenceTestStatistic". The test statistic, the linear statistic, and the transformed and original data structures.

estimates:

Object of class "list". The estimated parameters.

method:

Object of class "character". The test method.

call:

Object of class "call". The matched call.

Additionally, for objects of classes "ScalarIndependenceTest" or "ScalarIndependenceTestConfint":

parameter:

Object of class "character". The tested parameter.

nullvalue:

Object of class "numeric". The hypothesized value of the null hypothesis.

Additionally, for objects of class "ScalarIndependenceTestConfint":

confint:

Object of class "function". The confidence interval function.

conf.level:

Object of class "numeric". The confidence level.

Extends

For objects of classes "MaxTypeIndependenceTest", "QuadTypeIndependenceTest" or "ScalarIndependenceTest":
Class "IndependenceTest", directly.

For objects of class "ScalarIndependenceTestConfint":
Class "ScalarIndependenceTest", directly.
Class "IndependenceTest", by class "ScalarIndependenceTest", distance 2.

Known Subclasses

For objects of class "IndependenceTest":
Class "MaxTypeIndependenceTest", directly.
Class "QuadTypeIndependenceTest", directly.
Class "ScalarIndependenceTest", directly.
Class "ScalarIndependenceTestConfint", by class "ScalarIndependenceTest", distance 2.

For objects of class "ScalarIndependenceTest":
Class "ScalarIndependenceTestConfint", directly.

Methods

confint

signature(object = "IndependenceTest"): See the documentation for confint-methods (in package stats4) for details.

confint

signature(object = "ScalarIndependenceTestConfint"): See the documentation for confint-methods (in package stats4) for details.

covariance

signature(object = "IndependenceTest"): See the documentation for covariance() for details.

dperm

signature(object = "IndependenceTest"): See the documentation for dperm() for details.

expectation

signature(object = "IndependenceTest"): See the documentation for expectation() for details.

midpvalue

signature(object = "IndependenceTest"): See the documentation for midpvalue() for details.

pperm

signature(object = "IndependenceTest"): See the documentation for pperm() for details.

pvalue

signature(object = "IndependenceTest"): See the documentation for pvalue() for details.

pvalue

signature(object = "MaxTypeIndependenceTest"): See the documentation for pvalue() for details.

pvalue_interval

signature(object = "IndependenceTest"): See the documentation for pvalue_interval() for details.

qperm

signature(object = "IndependenceTest"): See the documentation for qperm() for details.

rperm

signature(object = "IndependenceTest"): See the documentation for rperm() for details.

show

signature(object = "IndependenceTest"): See the documentation for show() (in package methods) for details.

show

signature(object = "MaxTypeIndependenceTest"): See the documentation for show() (in package methods) for details.

show

signature(object = "QuadTypeIndependenceTest"): See the documentation for show() (in package methods) for details.

show

signature(object = "ScalarIndependenceTest"): See the documentation for show() (in package methods) for details.

show

signature(object = "ScalarIndependenceTestConfint"): See the documentation for show() (in package methods) for details.

size

signature(object = "IndependenceTest"): See the documentation for size() for details.

statistic

signature(object = "IndependenceTest"): See the documentation for statistic() for details.

support

signature(object = "IndependenceTest"): See the documentation for support() for details.

variance

signature(object = "IndependenceTest"): See the documentation for variance() for details.

---

Class "IndependenceTestProblem"

Description

Objects of class "IndependenceTestProblem" represent the transformed and original data structures corresponding to an independence problem.

Objects from the Class

Objects can be created by calls of the form

     new("IndependenceTestProblem", object, xtrafo = trafo, ytrafo = trafo, ...)

where object is an object of class "IndependenceProblem", xtrafo is the regression function $g(\mathbf{X})$ and ytrafo is the influence function $h(\mathbf{Y})$.

Slots

xtrans:

Object of class "matrix". The transformed x.

ytrans:

Object of class "matrix". The transformed y.

xtrafo:

Object of class "function". The regression function for x.

ytrafo:

Object of class "function". The influence function for y.

x:

Object of class "data.frame". The variables x.

y:

Object of class "data.frame". The variables y.

block:

Object of class "factor". The block structure.

weights:

Object of class "numeric". The case weights.

Extends

Class "IndependenceProblem", directly.

Known Subclasses

Class "IndependenceLinearStatistic", directly.
Class "IndependenceTestStatistic", by class "IndependenceLinearStatistic", distance 2.
Class "MaxTypeIndependenceTestStatistic", by class "IndependenceTestStatistic", distance 3.
Class "QuadTypeIndependenceTestStatistic", by class "IndependenceTestStatistic", distance 3.
Class "ScalarIndependenceTestStatistic", by class "IndependenceTestStatistic", distance 3.

Methods

initialize

signature(.Object = "IndependenceTestProblem"): See the documentation for initialize() (in package methods) for details.

---

Class "IndependenceTestStatistic" and Its Subclasses

Description

Objects of class "IndependenceTestStatistic" and its subclasses "MaxTypeIndependenceTestStatistic", "QuadTypeIndependenceTestStatistic" and "ScalarIndependenceTestStatistic" represent the test statistic, the linear statistic, and the transformed and original data structures corresponding to an independence problem.

Objects from the Class

Class "IndependenceTestStatistic" is a virtual class, so objects cannot be created from it directly.

Objects can be created by calls of the form

     new("MaxTypeIndependenceTestStatistic", object,
         alternative = c("two.sided", "less", "greater"), ...),

     new("QuadTypeIndependenceTestStatistic", object, paired = FALSE, ...)

and

     new("ScalarIndependenceTestStatistic", object,
         alternative = c("two.sided", "less", "greater"), paired = FALSE, ...)

where object is an object of class "IndependenceLinearStatistic", alternative is a character specifying the direction of the alternative hypothesis and paired is a logical indicating that paired data have been transformed in such a way that the (unstandardized) linear statistic is the sum of the absolute values of the positive differences between the paired observations.

Slots

For objects of classes "IndependenceTestStatistic", "MaxTypeIndependenceTestStatistic", "QuadTypeIndependenceTestStatistic" or "ScalarIndependenceTestStatistic":

teststatistic:

Object of class "numeric". The test statistic.

standardizedlinearstatistic:

Object of class "numeric". The standardized linear statistic.

linearstatistic:

Object of class "matrix". The linear statistic for each block.

expectation:

Object of class "matrix". The expectation of the linear statistic for each block.

covariance:

Object of class "matrix". The lower triangular elements of the covariance of the linear statistic for each block.

xtrans:

Object of class "matrix". The transformed x.

ytrans:

Object of class "matrix". The transformed y.

xtrafo:

Object of class "function". The regression function for x.

ytrafo:

Object of class "function". The influence function for y.

x:

Object of class "data.frame". The variables x.

y:

Object of class "data.frame". The variables y.

block:

Object of class "factor". The block structure.

weights:

Object of class "numeric". The case weights.

Additionally, for objects of classes "MaxTypeIndependenceTest" or "ScalarIndependenceTest":

alternative:

Object of class "character". The direction of the alternative hypothesis.

Additionally, for objects of class "QuadTypeIndependenceTest":

covarianceplus:

Object of class "numeric". The lower triangular elements of the Moore-Penrose inverse of the covariance of the linear statistic.

df:

Object of class "numeric". The rank of the covariance matrix.

Additionally, for objects of classes "QuadTypeIndependenceTest" or "ScalarIndependenceTest":

paired:

Object of class "logical". The indicator for paired test statistics.

Extends

For objects of class "IndependenceTestStatistic":
Class "IndependenceLinearStatistic", directly.
Class "IndependenceTestProblem", by class "IndependenceLinearStatistic", distance 2.
Class "IndependenceProblem", by class "IndependenceLinearStatistic", distance 3.

For objects of classes "MaxTypeIndependenceTestStatistic", "QuadTypeIndependenceTestStatistic" or "ScalarIndependenceTestStatistic":
Class "IndependenceTestStatistic", directly.
Class "IndependenceLinearStatistic", by class "IndependenceTestStatistic", distance 2.
Class "IndependenceTestProblem", by class "IndependenceTestStatistic", distance 3.
Class "IndependenceProblem", by class "IndependenceTestStatistic", distance 4.

Known Subclasses

For objects of class "IndependenceTestStatistic":
Class "MaxTypeIndependenceTestStatistic", directly.
Class "QuadTypeIndependenceTestStatistic", directly.
Class "ScalarIndependenceTestStatistic", directly.

Methods

ApproxNullDistribution

signature(object = "MaxTypeIndependenceTestStatistic"): See the documentation for ApproxNullDistribution() for details.

ApproxNullDistribution

signature(object = "QuadTypeIndependenceTestStatistic"): See the documentation for ApproxNullDistribution() for details.

ApproxNullDistribution

signature(object = "ScalarIndependenceTestStatistic"): See the documentation for ApproxNullDistribution() for details.

AsymptNullDistribution

signature(object = "MaxTypeIndependenceTestStatistic"): See the documentation for AsymptNullDistribution() for details.

AsymptNullDistribution

signature(object = "QuadTypeIndependenceTestStatistic"): See the documentation for AsymptNullDistribution() for details.

AsymptNullDistribution

signature(object = "ScalarIndependenceTestStatistic"): See the documentation for AsymptNullDistribution() for details.

ExactNullDistribution

signature(object = "QuadTypeIndependenceTestStatistic"): See the documentation for ExactNullDistribution() for details.

ExactNullDistribution

signature(object = "ScalarIndependenceTestStatistic"): See the documentation for ExactNullDistribution() for details.

covariance

signature(object = "QuadTypeIndependenceTestStatistic"): See the documentation for covariance() for details.

initialize

signature(.Object = "IndependenceTestStatistic"): See the documentation for initialize() (in package methods) for details.

initialize

signature(.Object = "MaxTypeIndependenceTestStatistic"): See the documentation for initialize() (in package methods) for details.

initialize

signature(.Object = "QuadTypeIndependenceTestStatistic"): See the documentation for initialize() (in package methods) for details.

initialize

signature(.Object = "ScalarIndependenceTestStatistic"): See the documentation for initialize() (in package methods) for details.

statistic

signature(object = "IndependenceTestStatistic"): See the documentation for statistic() for details.

---

Income and Job Satisfaction

Description

Income and job satisfaction by gender.

Usage

jobsatisfaction

Format

A contingency table with 104 observations on 3 variables.

Income

a factor with levels "<5000", "5000-15000", "15000-25000" and ">25000".

Job.Satisfaction

a factor with levels "Very Dissatisfied", "A Little Satisfied", "Moderately Satisfied" and "Very Satisfied".

Gender

a factor with levels "Female" and "Male".

Details

This data set was given in Agresti (2002, p. 288, Tab. 7.8). Winell and Lindbäck (2018) used the data to demonstrate a score-independent test for ordered categorical data.

Source

Agresti, A. (2002). Categorical Data Analysis, Second Edition. Hoboken, New Jersey: John Wiley & Sons.

References

Winell, H. and Lindbäck, J. (2018). A general score-independent test for order-restricted inference. Statistics in Medicine 37(21), 3078–3090. doi:10.1002/sim.7690

Examples

## Approximative (Monte Carlo) linear-by-linear association test
lbl_test(jobsatisfaction, distribution = approximate(nresample = 10000))

## Not run: 
## Approximative (Monte Carlo) score-independent test
## Winell and Lindbaeck (2018)
(it <- independence_test(jobsatisfaction,
                         distribution = approximate(nresample = 10000),
                         xtrafo = function(data)
                             trafo(data, factor_trafo = function(x)
                                 zheng_trafo(as.ordered(x))),
                         ytrafo = function(data)
                             trafo(data, factor_trafo = function(y)
                                 zheng_trafo(as.ordered(y)))))

## Extract the "best" set of scores
ss <- statistic(it, type = "standardized")
idx <- which(abs(ss) == max(abs(ss)), arr.ind = TRUE)
ss[idx[1], idx[2], drop = FALSE]
## End(Not run)

---

Two- and $K$-Sample Location Tests

Description

Testing the equality of the distributions of a numeric response variable in two or more independent groups against shift alternatives.

Usage

## S3 method for class 'formula'
oneway_test(formula, data, subset = NULL, weights = NULL, ...)
## S3 method for class 'IndependenceProblem'
oneway_test(object, ...)

## S3 method for class 'formula'
wilcox_test(formula, data, subset = NULL, weights = NULL, ...)
## S3 method for class 'IndependenceProblem'
wilcox_test(object, conf.int = FALSE, conf.level = 0.95, ...)

## S3 method for class 'formula'
kruskal_test(formula, data, subset = NULL, weights = NULL, ...)
## S3 method for class 'IndependenceProblem'
kruskal_test(object, ...)

## S3 method for class 'formula'
normal_test(formula, data, subset = NULL, weights = NULL, ...)
## S3 method for class 'IndependenceProblem'
normal_test(object, ties.method = c("mid-ranks", "average-scores"),
            conf.int = FALSE, conf.level = 0.95, ...)

## S3 method for class 'formula'
median_test(formula, data, subset = NULL, weights = NULL, ...)
## S3 method for class 'IndependenceProblem'
median_test(object, mid.score = c("0", "0.5", "1"),
            conf.int = FALSE, conf.level = 0.95, ...)

## S3 method for class 'formula'
savage_test(formula, data, subset = NULL, weights = NULL, ...)
## S3 method for class 'IndependenceProblem'
savage_test(object, ties.method = c("mid-ranks", "average-scores"),
            conf.int = FALSE, conf.level = 0.95, ...)

Arguments

formula	a formula of the form y ~ x | block where y is a numeric variable, x is a factor and block is an optional factor for stratification.
data	an optional data frame containing the variables in the model formula.
subset	an optional vector specifying a subset of observations to be used. Defaults to NULL.
weights	an optional formula of the form ~ w defining integer valued case weights for each observation. Defaults to NULL, implying equal weight for all observations.
object	an object inheriting from class "IndependenceProblem".
conf.int	a logical indicating whether a confidence interval for the difference in location should be computed. Defaults to FALSE.
conf.level	a numeric, confidence level of the interval. Defaults to 0.95.
ties.method	a character, the method used to handle ties: the score generating function either uses mid-ranks ("mid-ranks", default) or averages the scores of randomly broken ties ("average-scores").
mid.score	a character, the score assigned to observations exactly equal to the median: either 0 ("0", default), 0.5 ("0.5") or 1 ("1"); see ‘Details’.
...	further arguments to be passed to independence_test().

Details

oneway_test(), wilcox_test(), kruskal_test(), normal_test(), median_test() and savage_test() provide the Fisher-Pitman permutation test, the Wilcoxon-Mann-Whitney test, the Kruskal-Wallis test, the van der Waerden test, the Brown-Mood median test and the Savage test. A general description of these methods is given by Hollander and Wolfe (1999). For the adjustment of scores for tied values see Hájek, Šidák and Sen (1999, pp. 133–135).

The null hypothesis of equality, or conditional equality given block, of the distribution of y in the groups defined by x is tested against shift alternatives. In the two-sample case, the two-sided null hypothesis is $H_0\!: \mu = 0$, where $\mu = Y_1 - Y_2$ and $Y_s$ is the median of the responses in the $s$th sample. In case alternative = "less", the null hypothesis is $H_0\!: \mu \ge 0$. When alternative = "greater", the null hypothesis is $H_0\!: \mu \le 0$. Confidence intervals for the difference in location are available (except for oneway_test()) and computed according to Bauer (1972).

If x is an ordered factor, the default scores, 1:nlevels(x), can be altered using the scores argument (see independence_test()); this argument can also be used to coerce nominal factors to class "ordered". In this case, a linear-by-linear association test is computed and the direction of the alternative hypothesis can be specified using the alternative argument.

The Brown-Mood median test offers a choice of mid-score, i.e., the score assigned to observations exactly equal to the median. In the two-sample case, mid-score = "0" implies that the linear test statistic is simply the number of subjects in the second sample with observations greater than the median of the pooled sample. Similarly, the linear test statistic for the last alternative, mid-score = "1", is the number of subjects in the second sample with observations greater than or equal to the median of the pooled sample. If mid-score = "0.5" is selected, the linear test statistic is the mean of the test statistics corresponding to the first and last alternatives and has a symmetric distribution, or at least approximately so, under the null hypothesis (see Hájek, Šidák and Sen, 1999, pp. 97–98).

The conditional null distribution of the test statistic is used to obtain $p$-values and an asymptotic approximation of the exact distribution is used by default (distribution = "asymptotic"). Alternatively, the distribution can be approximated via Monte Carlo resampling or computed exactly for univariate two-sample problems by setting distribution to "approximate" or "exact", respectively. See asymptotic(), approximate() and exact() for details.

Value

An object inheriting from class "IndependenceTest". Confidence intervals can be extracted by confint().

Note

Starting with coin version 1.1-0, oneway_test() no longer allows the test statistic to be specified; a quadratic form is now used in the $K$-sample case. Please use independence_test() if more control is desired.

References

Bauer, D. F. (1972). Constructing confidence sets using rank statistics. Journal of the American Statistical Association 67(339), 687–690. doi:10.1080/01621459.1972.10481279

Hájek, J., Šidák, Z. and Sen, P. K. (1999). Theory of Rank Tests, Second Edition. San Diego: Academic Press.

Hollander, M. and Wolfe, D. A. (1999). Nonparametric Statistical Methods, Second Edition. New York: John Wiley & Sons.

Examples

## Tritiated Water Diffusion Across Human Chorioamnion
## Hollander and Wolfe (1999, p. 110, Tab. 4.1)
diffusion <- data.frame(
    pd = c(0.80, 0.83, 1.89, 1.04, 1.45, 1.38, 1.91, 1.64, 0.73, 1.46,
           1.15, 0.88, 0.90, 0.74, 1.21),
    age = factor(rep(c("At term", "12-26 Weeks"), c(10, 5)))
)

## Exact Wilcoxon-Mann-Whitney test
## Hollander and Wolfe (1999, p. 111)
## (At term - 12-26 Weeks)
(wt <- wilcox_test(pd ~ age, data = diffusion,
                   distribution = "exact", conf.int = TRUE))

## Extract observed Wilcoxon statistic
## Note: this is the sum of the ranks for age = "12-26 Weeks"
statistic(wt, type = "linear")

## Expectation, variance, two-sided pvalue and confidence interval
expectation(wt)
covariance(wt)
pvalue(wt)
confint(wt)

## For two samples, the Kruskal-Wallis test is equivalent to the W-M-W test
kruskal_test(pd ~ age, data = diffusion,
             distribution = "exact")

## Asymptotic Fisher-Pitman test
oneway_test(pd ~ age, data = diffusion)

## Approximative (Monte Carlo) Fisher-Pitman test
pvalue(oneway_test(pd ~ age, data = diffusion,
                   distribution = approximate(nresample = 10000)))

## Exact Fisher-Pitman test
pvalue(ot <- oneway_test(pd ~ age, data = diffusion,
                         distribution = "exact"))

## Plot density and distribution of the standardized test statistic
op <- par(no.readonly = TRUE) # save current settings
layout(matrix(1:2, nrow = 2))
s <- support(ot)
d <- dperm(ot, s)
p <- pperm(ot, s)
plot(s, d, type = "S", xlab = "Test Statistic", ylab = "Density")
plot(s, p, type = "S", xlab = "Test Statistic", ylab = "Cum. Probability")
par(op) # reset


## Example data
ex <- data.frame(
    y = c(3, 4, 8, 9, 1, 2, 5, 6, 7),
    x = factor(rep(c("no", "yes"), c(4, 5)))
)

## Boxplots
boxplot(y ~ x, data = ex)

## Exact Brown-Mood median test with different mid-scores
(mt1 <- median_test(y ~ x, data = ex, distribution = "exact"))
(mt2 <- median_test(y ~ x, data = ex, distribution = "exact",
                    mid.score = "0.5"))
(mt3 <- median_test(y ~ x, data = ex, distribution = "exact",
                    mid.score = "1")) # sign change!

## Plot density and distribution of the standardized test statistics
op <- par(no.readonly = TRUE) # save current settings
layout(matrix(1:3, nrow = 3))
s1 <- support(mt1); d1 <- dperm(mt1, s1)
plot(s1, d1, type = "h", main = "Mid-score: 0",
     xlab = "Test Statistic", ylab = "Density")
s2 <- support(mt2); d2 <- dperm(mt2, s2)
plot(s2, d2, type = "h", main = "Mid-score: 0.5",
     xlab = "Test Statistic", ylab = "Density")
s3 <- support(mt3); d3 <- dperm(mt3, s3)
plot(s3, d3, type = "h", main = "Mid-score: 1",
     xlab = "Test Statistic", ylab = "Density")
par(op) # reset


## Length of YOY Gizzard Shad
## Hollander and Wolfe (1999, p. 200, Tab. 6.3)
yoy <- data.frame(
    length = c(46, 28, 46, 37, 32, 41, 42, 45, 38, 44,
               42, 60, 32, 42, 45, 58, 27, 51, 42, 52,
               38, 33, 26, 25, 28, 28, 26, 27, 27, 27,
               31, 30, 27, 29, 30, 25, 25, 24, 27, 30),
    site = gl(4, 10, labels = as.roman(1:4))
)

## Approximative (Monte Carlo) Kruskal-Wallis test
kruskal_test(length ~ site, data = yoy,
             distribution = approximate(nresample = 10000))

## Approximative (Monte Carlo) Nemenyi-Damico-Wolfe-Dunn test (joint ranking)
## Hollander and Wolfe (1999, p. 244)
## (where Steel-Dwass results are given)
it <- independence_test(length ~ site, data = yoy,
                        distribution = approximate(nresample = 50000),
                        ytrafo = function(data)
                            trafo(data, numeric_trafo = rank_trafo),
                        xtrafo = mcp_trafo(site = "Tukey"))

## Global p-value
pvalue(it)

## Sites (I = II) != (III = IV) at alpha = 0.01 (p. 244)
pvalue(it, method = "single-step") # subset pivotality is violated

---

Maternal Drinking and Congenital Sex Organ Malformation

Description

A subset of data from a study on the relationship between maternal alcohol consumption and congenital malformations.

Usage

malformations

Format

A data frame with 32574 observations on 2 variables.

consumption

alcohol consumption, an ordered factor with levels "0", "<1", "1-2", "3-5" and ">=6".

malformation

congenital sex organ malformation, a factor with levels "Present" and "Absent".

Details

Data from a prospective study undertaken to determine whether moderate or light drinking during the first trimester of pregnancy increases the risk for congenital malformations (Mills and Graubard, 1987). The subset given here concerns only sex organ malformation (Mills and Graubard, 1987, Tab. 4).

This data set was used by Graubard and Korn (1987) to illustrate that different choices of scores for ordinal variables can lead to conflicting conclusions. Zheng (2008) also used the data, demonstrating two different score-independent tests for ordered categorical data; see also Winell and Lindbäck (2018).

Source

Mills, J. L. and Graubard, B. I. (1987). Is moderate drinking during pregnancy associated with an increased risk for malformations? Pediatrics 80(3), 309–314.

References

Graubard, B. I. and Korn, E. L. (1987). Choice of column scores for testing independence in ordered $2 \times K$ contingency tables. Biometrics 43(2), 471–476. doi:10.2307/2531828

Winell, H. and Lindbäck, J. (2018). A general score-independent test for order-restricted inference. Statistics in Medicine 37(21), 3078–3090. doi:10.1002/sim.7690

Zheng, G. (2008). Analysis of ordered categorical data: Two score-independent approaches. Biometrics 64(4), 1276–-1279. doi:10.1111/j.1541-0420.2008.00992.x

Examples

## Graubard and Korn (1987, Tab. 3)

## One-sided approximative (Monte Carlo) Cochran-Armitage test
## Note: midpoint scores (p < 0.05)
midpoints <- c(0, 0.5, 1.5, 4.0, 7.0)
chisq_test(malformation ~ consumption, data = malformations,
           distribution = approximate(nresample = 1000),
           alternative = "greater",
           scores = list(consumption = midpoints))

## One-sided approximative (Monte Carlo) Cochran-Armitage test
## Note: midrank scores (p > 0.05)
midranks <- c(8557.5, 24375.5, 32013.0, 32473.0, 32555.5)
chisq_test(malformation ~ consumption, data = malformations,
           distribution = approximate(nresample = 1000),
           alternative = "greater",
           scores = list(consumption = midranks))

## One-sided approximative (Monte Carlo) Cochran-Armitage test
## Note: equally spaced scores (p > 0.05)
chisq_test(malformation ~ consumption, data = malformations,
           distribution = approximate(nresample = 1000),
           alternative = "greater")

## Not run: 
## One-sided approximative (Monte Carlo) score-independent test
## Winell and Lindbaeck (2018)
(it <- independence_test(malformation ~ consumption, data = malformations,
                         distribution = approximate(nresample = 1000,
                                                    parallel = "snow",
                                                    ncpus = 8),
                         alternative = "greater",
                         xtrafo = function(data)
                             trafo(data, ordered_trafo = zheng_trafo)))

## Extract the "best" set of scores
ss <- statistic(it, type = "standardized")
idx <- which(ss == max(ss), arr.ind = TRUE)
ss[idx[1], idx[2], drop = FALSE]
## End(Not run)

---

Marginal Homogeneity Tests

Description

Testing the marginal homogeneity of a repeated measurements factor in a complete block design.

Usage

## S3 method for class 'formula'
mh_test(formula, data, subset = NULL, ...)
## S3 method for class 'table'
mh_test(object, ...)
## S3 method for class 'SymmetryProblem'
mh_test(object, ...)

Arguments

formula	a formula of the form y ~ x | block where y and x are factors and block is an optional factor (which is generated automatically if omitted).
data	an optional data frame containing the variables in the model formula.
subset	an optional vector specifying a subset of observations to be used. Defaults to NULL.
object	an object inheriting from classes "table" (with identical dimnames components) or "SymmetryProblem".
...	further arguments to be passed to symmetry_test().

Details

mh_test() provides the McNemar test, the Cochran $Q$ test, the Stuart(-Maxwell) test and the Madansky test of interchangeability. A general description of these methods is given by Agresti (2002).

The null hypothesis of marginal homogeneity is tested. The response variable and the measurement conditions are given by y and x, respectively, and block is a factor where each level corresponds to exactly one subject with repeated measurements.

This procedure is known as the McNemar test (McNemar, 1947) when both y and x are binary factors, as the Cochran $Q$ test (Cochran, 1950) when y is a binary factor and x is a factor with an arbitrary number of levels, as the Stuart(-Maxwell) test (Stuart, 1955; Maxwell, 1970) when y is a factor with an arbitrary number of levels and x is a binary factor, and as the Madansky test of interchangeability (Madansky, 1963), which implies marginal homogeneity, when both y and x are factors with an arbitrary number of levels.

If y and/or x are ordered factors, the default scores, 1:nlevels(y) and 1:nlevels(x), respectively, can be altered using the scores argument (see symmetry_test()); this argument can also be used to coerce nominal factors to class "ordered". If both y and x are ordered factors, a linear-by-linear association test is computed and the direction of the alternative hypothesis can be specified using the alternative argument. This extension was given by Birch (1965) who also discussed the situation when either the response or the measurement condition is an ordered factor; see also White, Landis and Cooper (1982).

The conditional null distribution of the test statistic is used to obtain $p$-values and an asymptotic approximation of the exact distribution is used by default (distribution = "asymptotic"). Alternatively, the distribution can be approximated via Monte Carlo resampling or computed exactly for univariate two-sample problems by setting distribution to "approximate" or "exact", respectively. See asymptotic(), approximate() and exact() for details.

Value

An object inheriting from class "IndependenceTest".

Note

This function is currently computationally inefficient for data with a large number of pairs or sets.

References

Agresti, A. (2002). Categorical Data Analysis, Second Edition. Hoboken, New Jersey: John Wiley & Sons.

Birch, M. W. (1965). The detection of partial association, II: The general case. Journal of the Royal Statistical Society B 27(1), 111–124. doi:10.1111/j.2517-6161.1965.tb00593.x

Cochran, W. G. (1950). The comparison of percentages in matched samples. Biometrika 37(3/4), 256–266. doi:10.1093/biomet/37.3-4.256

Madansky, A. (1963). Tests of homogeneity for correlated samples. Journal of the American Statistical Association 58(301), 97–119. doi:10.1080/01621459.1963.10500835

Maxwell, A. E. (1970). Comparing the classification of subjects by two independent judges. British Journal of Psychiatry 116(535), 651–655. doi:10.1192/bjp.116.535.651

McNemar, Q. (1947). Note on the sampling error of the difference between correlated proportions or percentages. Psychometrika 12(2), 153–157. doi:10.1007/BF02295996

Stuart, A. (1955). A test for homogeneity of the marginal distributions in a two-way classification. Biometrika 42(3/4), 412–416. doi:10.1093/biomet/42.3-4.412

White, A. A., Landis, J. R. and Cooper, M. M. (1982). A note on the equivalence of several marginal homogeneity test criteria for categorical data. International Statistical Review 50(1), 27–34. doi:10.2307/1402457

Examples

## Performance of prime minister
## Agresti (2002, p. 409)
performance <- matrix(
    c(794, 150,
       86, 570),
    nrow = 2, byrow = TRUE,
    dimnames = list(
         "First" = c("Approve", "Disprove"),
        "Second" = c("Approve", "Disprove")
    )
)
performance <- as.table(performance)
diag(performance) <- 0 # speed-up: only off-diagonal elements contribute

## Asymptotic McNemar Test
mh_test(performance)

## Exact McNemar Test
mh_test(performance, distribution = "exact")


## Effectiveness of different media for the growth of diphtheria
## Cochran (1950, Tab. 2)
cases <- c(4, 2, 3, 1, 59)
n <- sum(cases)
cochran <- data.frame(
    diphtheria = factor(
        unlist(rep(list(c(1, 1, 1, 1),
                        c(1, 1, 0, 1),
                        c(0, 1, 1, 1),
                        c(0, 1, 0, 1),
                        c(0, 0, 0, 0)),
                   cases))
    ),
    media = factor(rep(LETTERS[1:4], n)),
    case =  factor(rep(seq_len(n), each = 4))
)

## Asymptotic Cochran Q test (Cochran, 1950, p. 260)
mh_test(diphtheria ~ media | case, data = cochran) # Q = 8.05

## Approximative Cochran Q test
mt <- mh_test(diphtheria ~ media | case, data = cochran,
              distribution = approximate(nresample = 10000))
pvalue(mt)             # standard p-value
midpvalue(mt)          # mid-p-value
pvalue_interval(mt)    # p-value interval
size(mt, alpha = 0.05) # test size at alpha = 0.05 using the p-value


## Opinions on Pre- and Extramarital Sex
## Agresti (2002, p. 421)
opinions <- c("Always wrong", "Almost always wrong",
              "Wrong only sometimes", "Not wrong at all")
PreExSex <- matrix(
    c(144, 33, 84, 126,
        2,  4, 14,  29,
        0,  2,  6,  25,
        0,  0,  1,   5),
    nrow = 4,
    dimnames = list(
          "Premarital Sex" = opinions,
        "Extramarital Sex" = opinions
    )
)
PreExSex <- as.table(PreExSex)

## Asymptotic Stuart test
mh_test(PreExSex)

## Asymptotic Stuart-Birch test
## Note: response as ordinal
mh_test(PreExSex, scores = list(response = 1:length(opinions)))


## Vote intention
## Madansky (1963, pp. 107-108)
vote <- array(
    c(120, 1,  8, 2,   2,  1, 2, 1,  7,
        6, 2,  1, 1, 103,  5, 1, 4,  8,
       20, 3, 31, 1,   6, 30, 2, 1, 81),
    dim = c(3, 3, 3),
    dimnames = list(
          "July" = c("Republican", "Democratic", "Uncertain"),
        "August" = c("Republican", "Democratic", "Uncertain"),
          "June" = c("Republican", "Democratic", "Uncertain")
    )
)
vote <- as.table(vote)

## Asymptotic Madansky test (Q = 70.77)
mh_test(vote)


## Cross-over study
## http://www.nesug.org/proceedings/nesug00/st/st9005.pdf
dysmenorrhea <- array(
    c(6, 2, 1,  3, 1, 0,  1, 2, 1,
      4, 3, 0, 13, 3, 0,  8, 1, 1,
      5, 2, 2, 10, 1, 0, 14, 2, 0),
    dim = c(3, 3, 3),
    dimnames =  list(
          "Placebo" = c("None", "Moderate", "Complete"),
         "Low dose" = c("None", "Moderate", "Complete"),
        "High dose" = c("None", "Moderate", "Complete")
    )
)
dysmenorrhea <- as.table(dysmenorrhea)

## Asymptotic Madansky-Birch test (Q = 53.76)
## Note: response as ordinal
mh_test(dysmenorrhea, scores = list(response = 1:3))

## Asymptotic Madansky-Birch test (Q = 47.29)
## Note: response and measurement conditions as ordinal
mh_test(dysmenorrhea, scores = list(response = 1:3,
                                    conditions = 1:3))

---

Generalized Maximally Selected Statistics

Description

Testing the independence of two sets of variables measured on arbitrary scales against cutpoint alternatives.

Usage

## S3 method for class 'formula'
maxstat_test(formula, data, subset = NULL, weights = NULL, ...)
## S3 method for class 'table'
maxstat_test(object, ...)
## S3 method for class 'IndependenceProblem'
maxstat_test(object, teststat = c("maximum", "quadratic"),
             distribution = c("asymptotic", "approximate", "none"),
             minprob = 0.1, maxprob = 1 - minprob, ...)

Arguments

formula	a formula of the form y1 + ... + yq ~ x1 + ... + xp | block where y1, ..., yq and x1, ..., xp are measured on arbitrary scales (nominal, ordinal or continuous with or without censoring) and block is an optional factor for stratification.
data	an optional data frame containing the variables in the model formula.
subset	an optional vector specifying a subset of observations to be used. Defaults to NULL.
weights	an optional formula of the form ~ w defining integer valued case weights for each observation. Defaults to NULL, implying equal weight for all observations.
object	an object inheriting from classes "table" or "IndependenceProblem".
teststat	a character, the type of test statistic to be applied: either a maximum statistic ("maximum", default) or a quadratic form ("quadratic").
distribution	a character, the conditional null distribution of the test statistic can be approximated by its asymptotic distribution ("asymptotic", default) or via Monte Carlo resampling ("approximate"). Alternatively, the functions asymptotic or approximate can be used. Computation of the null distribution can be suppressed by specifying "none".
minprob	a numeric, a fraction between 0 and 0.5 specifying that cutpoints only greater than the minprob $\cdot$ 100% quantile of x1, ..., xp are considered. Defaults to 0.1.
maxprob	a numeric, a fraction between 0.5 and 1 specifying that cutpoints only smaller than the maxprob $\cdot$ 100% quantile of x1, ..., xp are considered. Defaults to 1 - minprob.
...	further arguments to be passed to independence_test().

Details

maxstat_test() provides generalized maximally selected statistics. The family of maximally selected statistics encompasses a large collection of procedures used for the estimation of simple cutpoint models including, but not limited to, maximally selected $\chi^2$ statistics, maximally selected Cochran-Armitage statistics, maximally selected rank statistics and maximally selected statistics for multiple covariates. A general description of these methods is given by Hothorn and Zeileis (2008).

The null hypothesis of independence, or conditional independence given block, between y1, ..., yq and x1, ..., xp is tested against cutpoint alternatives. All possible partitions into two groups are evaluated for each unordered covariate x1, ..., xp, whereas only order-preserving binary partitions are evaluated for ordered or numeric covariates. The cutpoint is then a set of levels defining one of the two groups.

If both response and covariate is univariable, say y1 and x1, this procedure is known as maximally selected $\chi^2$ statistics (Miller and Siegmund, 1982) when y1 is a binary factor and x1 is a numeric variable, and as maximally selected rank statistics when y1 is a rank transformed numeric variable and x1 is a numeric variable (Lausen and Schumacher, 1992). Lausen et al. (2004) introduced maximally selected statistics for a univariable numeric response and multiple numeric covariates x1, ..., xp.

If, say, y1 and/or x1 are ordered factors, the default scores, 1:nlevels(y1) and 1:nlevels(x1), respectively, can be altered using the scores argument (see independence_test()); this argument can also be used to coerce nominal factors to class "ordered". If both, say, y1 and x1 are ordered factors, a linear-by-linear association test is computed and the direction of the alternative hypothesis can be specified using the alternative argument. The particular extension to the case of a univariable ordered response and a univariable numeric covariate was given by Betensky and Rabinowitz (1999) and is known as maximally selected Cochran-Armitage statistics.

The conditional null distribution of the test statistic is used to obtain $p$-values and an asymptotic approximation of the exact distribution is used by default (distribution = "asymptotic"). Alternatively, the distribution can be approximated via Monte Carlo resampling by setting distribution to "approximate". See asymptotic() and approximate() for details.

Value

An object inheriting from class "IndependenceTest".

Note

Starting with coin version 1.1-0, maximum statistics and quadratic forms can no longer be specified using teststat = "maxtype" and teststat = "quadtype", respectively (as was used in versions prior to 0.4-5).

References

Betensky, R. A. and Rabinowitz, D. (1999). Maximally selected $\chi^2$ statistics for $k \times 2$ tables. Biometrics 55(1), 317–320. doi:10.1111/j.0006-341X.1999.00317.x

Hothorn, T. and Lausen, B. (2003). On the exact distribution of maximally selected rank statistics. Computational Statistics & Data Analysis 43(2), 121–137. doi:10.1016/S0167-9473(02)00225-6

Hothorn, T. and Zeileis, A. (2008). Generalized maximally selected statistics. Biometrics 64(4), 1263–1269. doi:10.1111/j.1541-0420.2008.00995.x

Lausen, B., Hothorn, T., Bretz, F. and Schumacher, M. (2004). Assessment of optimal selected prognostic factors. Biometrical Journal 46(3), 364–374. doi:10.1002/bimj.200310030

Lausen, B. and Schumacher, M. (1992). Maximally selected rank statistics. Biometrics 48(1), 73–85. doi:10.2307/2532740

Miller, R. and Siegmund, D. (1982). Maximally selected chi square statistics. Biometrics 38(4), 1011–1016. doi:10.2307/2529881

Müller, J. and Hothorn, T. (2004). Maximally selected two-sample statistics as a new tool for the identification and assessment of habitat factors with an application to breeding bird communities in oak forests. European Journal of Forest Research 123(3), 219–228. doi:10.1007/s10342-004-0035-5

Examples

## Tree pipit data (Mueller and Hothorn, 2004)
## Asymptotic maximally selected statistics
maxstat_test(counts ~ coverstorey, data = treepipit)

## Asymptotic maximally selected statistics
## Note: all covariates simultaneously
mt <- maxstat_test(counts ~ ., data = treepipit)
mt@estimates$estimate


## Malignant arrythmias data (Hothorn and Lausen, 2003, Sec. 7.2)
## Asymptotic maximally selected statistics
maxstat_test(Surv(time, event) ~  EF, data = hohnloser,
             ytrafo = function(data)
                 trafo(data, surv_trafo = function(y)
                     logrank_trafo(y, ties.method = "Hothorn-Lausen")))


## Breast cancer data (Hothorn and Lausen, 2003, Sec. 7.3)
## Asymptotic maximally selected statistics
data("sphase", package = "TH.data")
maxstat_test(Surv(RFS, event) ~  SPF, data = sphase,
             ytrafo = function(data)
                 trafo(data, surv_trafo = function(y)
                     logrank_trafo(y, ties.method = "Hothorn-Lausen")))


## Job satisfaction data (Agresti, 2002, p. 288, Tab. 7.8)
## Asymptotic maximally selected statistics
maxstat_test(jobsatisfaction)

## Asymptotic maximally selected statistics
## Note: 'Job.Satisfaction' and 'Income' as ordinal
maxstat_test(jobsatisfaction,
             scores = list("Job.Satisfaction" = 1:4,
                           "Income" = 1:4))

---

Chromosomal Effects of Mercury-Contaminated Fish Consumption

Description

The mercury level in blood, the proportion of cells with abnormalities, and the proportion of cells with chromosome aberrations in consumers of mercury-contaminated fish and a control group.

Usage

mercuryfish

Format

A data frame with 39 observations on 4 variables.

group

a factor with levels "control" and "exposed".

mercury

mercury level in blood.

abnormal

the proportion of cells with structural abnormalities.

ccells

the proportion of $C_u$ cells, i.e., cells with asymmetrical or incomplete-symmetrical chromosome aberrations.

Details

Control subjects ("control") and subjects who ate contaminated fish for more than three years ("exposed") are under study.

Rosenbaum (1994) proposed a coherence criterion defining a partial ordering, i.e., an observation is smaller than another when all responses are smaller, and a score reflecting the “ranking” is attached to each observation. The corresponding partially ordered set (POSET) test can be used to test if the distribution of the scores differ between the groups. Alternatively, a multivariate test can be applied.

Source

Skerfving, S., Hansson, K., Mangs, C., Lindsten, J. and Ryman, N. (1974). Methylmercury-induced chromosome damage in men. Environmental Research 7(1), 83–98. doi:10.1016/0013-9351(74)90078-4

References

Hothorn, T., Hornik, K., van de Wiel, M. A. and Zeileis, A. (2006). A Lego system for conditional inference. The American Statistician 60(3), 257–263. doi:10.1198/000313006X118430

Rosenbaum, P. R. (1994). Coherence in observational studies. Biometrics 50(2), 368–374. doi:10.2307/2533380

Examples

## Coherence criterion
coherence <- function(data) {
    x <- as.matrix(data)
    matrix(apply(x, 1, function(y)
        sum(colSums(t(x) < y) == ncol(x)) -
            sum(colSums(t(x) > y) == ncol(x))), ncol = 1)
}

## Asymptotic POSET test
poset <- independence_test(mercury + abnormal + ccells ~ group,
                           data = mercuryfish, ytrafo = coherence)

## Linear statistic (T in the notation of Rosenbaum, 1994)
statistic(poset, type = "linear")

## Expectation
expectation(poset)

## Variance
## Note: typo in Rosenbaum (1994, p. 371, Sec. 2, last paragraph)
variance(poset)

## Standardized statistic
statistic(poset)

## P-value
pvalue(poset)

## Exact POSET test
independence_test(mercury + abnormal + ccells ~ group,
                  data = mercuryfish, ytrafo = coherence,
                  distribution = "exact")

## Asymptotic multivariate test
mvtest <- independence_test(mercury + abnormal + ccells ~ group,
                            data = mercuryfish)

## Global p-value
pvalue(mvtest)

## Single-step adjusted p-values
pvalue(mvtest, method = "single-step")

## Step-down adjusted p-values
pvalue(mvtest, method = "step-down")

---

Acute Painful Diabetic Neuropathy

Description

The logarithm of the ratio of pain scores measured at baseline and after four weeks in a control group and a treatment group.

Usage

neuropathy

Format

A data frame with 58 observations on 2 variables.

pain

pain scores: ln(baseline / final).

group

a factor with levels "control" and "treat".

Details

Data from Conover and Salsburg (1988, Tab. 1).

Source

Conover, W. J. and Salsburg, D. S. (1988). Locally most powerful tests for detecting treatment effects when only a subset of patients can be expected to “respond” to treatment. Biometrics 44(1), 189–196. doi:10.2307/2531906

Examples

## Conover and Salsburg (1988, Tab. 2)

## One-sided approximative Fisher-Pitman test
oneway_test(pain ~ group, data = neuropathy,
            alternative = "less",
            distribution = approximate(nresample = 10000))

## One-sided approximative Wilcoxon-Mann-Whitney test
wilcox_test(pain ~ group, data = neuropathy,
            alternative = "less",
            distribution = approximate(nresample = 10000))

## One-sided approximative Conover-Salsburg test
oneway_test(pain ~ group, data = neuropathy,
            alternative = "less",
            distribution = approximate(nresample = 10000),
            ytrafo = function(data)
                trafo(data, numeric_trafo = consal_trafo))

## One-sided approximative maximum test for a range of 'a' values
it <- independence_test(pain ~ group, data = neuropathy,
                        alternative = "less",
                        distribution = approximate(nresample = 10000),
                        ytrafo = function(data)
                            trafo(data, numeric_trafo = function(y)
                                consal_trafo(y, a = 2:7)))
pvalue(it, method = "single-step")

---

Specification of the Reference Distribution

Description

Specification of the asymptotic, approximative (Monte Carlo) and exact reference distribution.

Usage

asymptotic(maxpts = 25000, abseps = 0.001, releps = 0)
approximate(nresample = 10000L, parallel = c("no", "multicore", "snow"),
            ncpus = 1L, cl = NULL, B)
exact(algorithm = c("auto", "shift", "split-up"), fact = NULL)

Arguments

maxpts	an integer, the maximum number of function values. Defaults to 25000.
abseps	a numeric, the absolute error tolerance. Defaults to 0.001.
releps	a numeric, the relative error tolerance. Defaults to 0.
nresample	a positive integer, the number of Monte Carlo replicates used for the computation of the approximative reference distribution. Defaults to 10000L.
parallel	a character, the type of parallel operation: either "no" (default), "multicore" or "snow".
ncpus	an integer, the number of processes to be used in parallel operation. Defaults to 1L.
cl	an object inheriting from class "cluster", specifying an optional parallel or snow cluster if parallel = "snow". Defaults to NULL.
B	deprecated, use nresample instead.
algorithm	a character, the algorithm used for the computation of the exact reference distribution: either "auto" (default), "shift" or "split-up".
fact	an integer to multiply the response values with. Defaults to NULL.

Details

asymptotic(), approximate() and exact() can be supplied to the distribution argument of, e.g., independence_test() to provide control of the specification of the asymptotic, approximative (Monte Carlo) and exact reference distribution, respectively.

The asymptotic reference distribution is computed using a randomised quasi-Monte Carlo method (Genz and Bretz, 2009) and is applicable to arbitrary covariance structures with dimensions up to 1000. See GenzBretz() in package mvtnorm for details on maxpts, abseps and releps.

The approximative (Monte Carlo) reference distribution is obtained by a conditional Monte Carlo procedure, i.e., by computing the test statistic for nresample random samples from all admissible permutations of the response $\bf{Y}$ within each block (Hothorn et al., 2008). By default, the distribution is computed using serial operation (parallel = "no"). The use of parallel operation is specified by setting parallel to either "multicore" (not available for MS Windows) or "snow". In the latter case, if cl = NULL (default) a cluster with ncpus processes is created on the local machine unless a default cluster has been registered (see setDefaultCluster() in package parallel) in which case that gets used instead. Alternatively, the use of an optional parallel or snow cluster can be specified by cl. See ‘Examples’ and package parallel for details on parallel operation.

The exact reference distribution, currently available for univariate two-sample problems only, is computed using either the shift algorithm (Streitberg and Röhmel, 1984, 1986, 1987) or the split-up algorithm (van de Wiel, 2001). The shift algorithm handles blocks pertaining to, e.g., pre- and post-stratification, but can only be used with positive integer-valued scores $h(\bf{Y})$. The split-up algorithm can be used with non-integer scores, but does not handle blocks. By default, an automatic choice is made (algorithm = "auto") but the shift and split-up algorithms can be selected by setting algorithm to "shift" or "split-up", respectively.

Note

Starting with coin version 1.1-0, the default for algorithm is "auto", having identical behaviour to "shift" in previous versions. In earlier versions of the package, algorithm = "shift" silently switched to the split-up algorithm if non-integer scores were detected, whereas the current version exits with a warning.

In versions of coin prior to 1.3-0, the number of Monte Carlo replicates in approximate() was specified using the now deprecated B argument. This will be made defunct and removed in a future release. It has been replaced by the nresample argument (for conformity with the libcoin, party and partykit packages).

References

Genz, A. and Bretz, F. (2009). Computation of Multivariate Normal and t Probabilities. Heidelberg: Springer-Verlag.

Hothorn, T., Hornik, K., van de Wiel, M. A. and Zeileis, A. (2008). Implementing a class of permutation tests: The coin package. Journal of Statistical Software 28(8), 1–23. doi:10.18637/jss.v028.i08

Streitberg, B. and Röhmel, J. (1984). Exact nonparametrics in APL. APL Quote Quad 14(4), 313–325. doi:10.1145/384283.801115

Streitberg, B. and Röhmel, J. (1986). Exact distributions for permutations and rank tests: an introduction to some recently published algorithms. Statistical Software Newsletter 12(1), 10–17.

Streitberg, B. and Röhmel, J. (1987). Exakte verteilungen für rang- und randomisierungstests im allgemeinen c-stichprobenfall. EDV in Medizin und Biologie 18(1), 12–19.

van de Wiel, M. A. (2001). The split-up algorithm: a fast symbolic method for computing p-values of distribution-free statistics. Computational Statistics 16(4), 519–538. doi:10.1007/s180-001-8328-6

Examples

## Approximative (Monte Carlo) Cochran-Mantel-Haenszel test

## Serial operation
set.seed(123)
cmh_test(disease ~ smoking | gender, data = alzheimer,
         distribution = approximate(nresample = 100000))

## Not run: 
## Multicore with 8 processes (not for MS Windows)
set.seed(123, kind = "L'Ecuyer-CMRG")
cmh_test(disease ~ smoking | gender, data = alzheimer,
         distribution = approximate(nresample = 100000,
                                    parallel = "multicore", ncpus = 8))

## Automatic PSOCK cluster with 4 processes
set.seed(123, kind = "L'Ecuyer-CMRG")
cmh_test(disease ~ smoking | gender, data = alzheimer,
         distribution = approximate(nresample = 100000,
                                    parallel = "snow", ncpus = 4))

## Registered FORK cluster with 12 processes (not for MS Windows)
fork12 <- parallel::makeCluster(12, "FORK") # set-up cluster
parallel::setDefaultCluster(fork12) # register default cluster
set.seed(123, kind = "L'Ecuyer-CMRG")
cmh_test(disease ~ smoking | gender, data = alzheimer,
         distribution = approximate(nresample = 100000,
                                    parallel = "snow"))
parallel::stopCluster(fork12) # clean-up

## User-specified PSOCK cluster with 8 processes
psock8 <- parallel::makeCluster(8, "PSOCK") # set-up cluster
set.seed(123, kind = "L'Ecuyer-CMRG")
cmh_test(disease ~ smoking | gender, data = alzheimer,
         distribution = approximate(nresample = 100000,
                                    parallel = "snow", cl = psock8))
parallel::stopCluster(psock8) # clean-up
## End(Not run)

---

Class "NullDistribution" and Its Subclasses

Description

Objects of class "NullDistribution" and its subclasses "ApproxNullDistribution", "AsymptNullDistribution" and "ExactNullDistribution" represent the reference distribution.

Objects from the Class

Objects can be created by calls of the form

     new("NullDistribution", ...),

     new("ApproxNullDistribution", ...),

     new("AsymptNullDistribution", ...)

and

     new("ExactNullDistribution", ...).

Slots

For objects of classes "NullDistribution", "ApproxNullDistribution", "AsymptNullDistribution" or "ExactNullDistribution":

name:

Object of class "character". The name of the reference distribution.

p:

Object of class "function". The distribution function of the reference distribution.

pvalue:

Object of class "function". The $p$-value function of the reference distribution.

parameters:

Object of class "list". Additional parameters.

support:

Object of class "function". The support of the reference distribution.

d:

Object of class "function". The density function of the reference distribution.

q:

Object of class "function". The quantile function of the reference distribution.

midpvalue:

Object of class "function". The mid-$p$-value function of the reference distribution.

pvalueinterval:

Object of class "function". The $p$-value interval function of the reference distribution.

size:

Object of class "function". The size function of the reference distribution.

Additionally, for objects of classes "ApproxNullDistribution" or "AsymptNullDistribution":

seed:

Object of class "integer". The random number generator state (i.e., the value of .Random.seed).

Additionally, for objects of class "ApproxNullDistribution":

nresample:

Object of class "numeric". The number of Monte Carlo replicates.

Extends

For objects of class "NullDistribution":
Class "PValue", directly.

For objects of classes "ApproxNullDistribution", "AsymptNullDistribution" or "ExactNullDistribution":
Class "NullDistribution", directly.
Class "PValue", by class "NullDistribution", distance 2.

Known Subclasses

For objects of class "NullDistribution":
Class "ApproxNullDistribution", directly.
Class "AsymptNullDistribution", directly.
Class "ExactNullDistribution", directly.

Methods

dperm

signature(object = "NullDistribution"): See the documentation for dperm() for details.

midpvalue

signature(object = "NullDistribution"): See the documentation for midpvalue() for details.

midpvalue

signature(object = "ApproxNullDistribution"): See the documentation for midpvalue() for details.

pperm

signature(object = "NullDistribution"): See the documentation for pperm() for details.

pvalue

signature(object = "NullDistribution"): See the documentation for pvalue() for details.

pvalue

signature(object = "ApproxNullDistribution"): See the documentation for pvalue() for details.

pvalue_interval

signature(object = "NullDistribution"): See the documentation for pvalue_interval() for details.

qperm

signature(object = "NullDistribution"): See the documentation for qperm() for details.

rperm

signature(object = "NullDistribution"): See the documentation for rperm() for details.

size

signature(object = "NullDistribution"): See the documentation for size() for details.

support

signature(object = "NullDistribution"): See the documentation for support() for details.

---

Computation of the Reference Distribution

Description

Methods for computation of the asymptotic, approximative (Monte Carlo) and exact reference distribution.

Usage

## S4 method for signature 'MaxTypeIndependenceTestStatistic'
AsymptNullDistribution(object, ...)
## S4 method for signature 'QuadTypeIndependenceTestStatistic'
AsymptNullDistribution(object, ...)
## S4 method for signature 'ScalarIndependenceTestStatistic'
AsymptNullDistribution(object, ...)

## S4 method for signature 'MaxTypeIndependenceTestStatistic'
ApproxNullDistribution(object, nresample = 10000L, B, ...)
## S4 method for signature 'QuadTypeIndependenceTestStatistic'
ApproxNullDistribution(object, nresample = 10000L, B, ...)
## S4 method for signature 'ScalarIndependenceTestStatistic'
ApproxNullDistribution(object, nresample = 10000L, B, ...)

## S4 method for signature 'QuadTypeIndependenceTestStatistic'
ExactNullDistribution(object, algorithm = c("auto", "shift", "split-up"), ...)
## S4 method for signature 'ScalarIndependenceTestStatistic'
ExactNullDistribution(object, algorithm = c("auto", "shift", "split-up"), ...)

Arguments

object	an object from which the asymptotic, approximative (Monte Carlo) or exact reference distribution can be computed.
nresample	a positive integer, the number of Monte Carlo replicates used for the computation of the approximative reference distribution. Defaults to 10000L.
B	deprecated, use nresample instead.
algorithm	a character, the algorithm used for the computation of the exact reference distribution: either "auto" (default), "shift" or "split-up".
...	further arguments to be passed to or from methods.

Details

The methods AsymptNullDistribution, ApproxNullDistribution and ExactNullDistribution compute the asymptotic, approximative (Monte Carlo) and exact reference distribution, respectively.

Value

An object of class "AsymptNullDistribution", "ApproxNullDistribution" or "ExactNullDistribution".

Note

In versions of coin prior to 1.3-0, the number of Monte Carlo replicates in ApproxNullDistribution() was specified using the now deprecated B argument. This will be made defunct and removed in a future release. It has been replaced by the nresample argument (for conformity with the libcoin, party and partykit packages).

---

Ovarian Carcinoma

Description

Survival times of 35 women suffering from ovarian carcinoma at stadium II and IIA.

Usage

ocarcinoma

Format

A data frame with 35 observations on 3 variables.

time

time (days).

stadium

a factor with levels "II" and "IIA".

event

status indicator for time: FALSE for right-censored observations and TRUE otherwise.

Details

Data from Fleming et al. (1980) and Fleming, Green and Harrington (1984). Reanalysed in Schumacher and Schulgen (2002).

Source

Fleming, T. R., Green, S. J. and Harrington, D. P. (1984). Considerations for monitoring and evaluating treatment effects in clinical trials. Controlled Clinical Trials 5(1), 55–66. doi:10.1016/0197-2456(84)90150-8

Fleming, T. R., O'Fallon, J. R., O'Brien, P. C. and Harrington, D. P. (1980). Modified Kolmogorov-Smirnov test procedures with application to arbitrarily right-censored data. Biometrics 36(4), 607–625. doi:10.2307/2556114

References

Schumacher, M. and Schulgen, G. (2002). Methodik Klinischer Studien: Methodische Grundlagen der Planung, Durchführung und Auswertung. Heidelberg: Springer.

Examples

## Exact logrank test
lt <- logrank_test(Surv(time, event) ~ stadium, data = ocarcinoma,
                   distribution = "exact")

## Test statistic
statistic(lt)

## P-value
pvalue(lt)

---

Computation of the Permutation Distribution

Description

Methods for computation of the density function, distribution function, quantile function, random numbers and support of the permutation distribution.

Usage

## S4 method for signature 'NullDistribution'
dperm(object, x, ...)
## S4 method for signature 'IndependenceTest'
dperm(object, x, ...)

## S4 method for signature 'NullDistribution'
pperm(object, q, ...)
## S4 method for signature 'IndependenceTest'
pperm(object, q, ...)

## S4 method for signature 'NullDistribution'
qperm(object, p, ...)
## S4 method for signature 'IndependenceTest'
qperm(object, p, ...)

## S4 method for signature 'NullDistribution'
rperm(object, n, ...)
## S4 method for signature 'IndependenceTest'
rperm(object, n, ...)

## S4 method for signature 'NullDistribution'
support(object, ...)
## S4 method for signature 'IndependenceTest'
support(object, ...)

Arguments

object	an object from which the density function, distribution function, quantile function, random numbers or support of the permutation distribution can be computed.
x, q	a numeric vector, the quantiles for which the density function or distribution function is computed.
p	a numeric vector, the probabilities for which the quantile function is computed.
n	a numeric vector, the number of observations. If length(n) > 1, the length is taken to be the number required.
...	further arguments to be passed to methods.

Details

The methods dperm, pperm, qperm, rperm and support compute the density function, distribution function, quantile function, random deviates and support, respectively, of the permutation distribution.

Value

The density function, distribution function, quantile function, random deviates or support of the permutation distribution computed from object. A numeric vector.

Note

The density of asymptotic permutation distributions for maximum-type tests or exact permutation distributions obtained by the split-up algorithm is reported as NA. The quantile function of asymptotic permutation distributions for maximum-type tests cannot be computed for p less than 0.5, due to limitations in the mvtnorm package. The support of exact permutation distributions obtained by the split-up algorithm is reported as NA.

In versions of coin prior to 1.1-0, the support of asymptotic permutation distributions was given as an interval containing 99.999 % of the probability mass. It is now reported as NA.

Examples

## Two-sample problem
dta <- data.frame(
    y = rnorm(20),
    x = gl(2, 10)
)

## Exact Ansari-Bradley test
at <- ansari_test(y ~ x, data = dta, distribution = "exact")

## Support of the exact distribution of the Ansari-Bradley statistic
supp <- support(at)

## Density of the exact distribution of the Ansari-Bradley statistic
dens <- dperm(at, x = supp)

## Plotting the density
plot(supp, dens, type = "s")

## 95% quantile
qperm(at, p = 0.95)

## One-sided p-value
pperm(at, q = statistic(at))

## Random number generation
rperm(at, n = 5)

---

Multiple Dosing Photococarcinogenicity Experiment

Description

Survival time, time to first tumor, and total number of tumors in three groups of animals in a photococarcinogenicity study.

Usage

photocar

Format

A data frame with 108 observations on 6 variables.

group

a factor with levels "A", "B", and "C".

ntumor

total number of tumors.

time

survival time.

event

status indicator for time: FALSE for right-censored observations and TRUE otherwise.

dmin

time to first tumor.

tumor

status indicator for dmin: FALSE when no tumor was observed and TRUE otherwise.

Details

The animals were exposed to different levels of ultraviolet radiation (UVR) exposure (group A: topical vehicle and 600 Robertson–Berger units of UVR, group B: no topical vehicle and 600 Robertson–Berger units of UVR and group C: no topical vehicle and 1200 Robertson–Berger units of UVR). The data are taken from Tables 1 to 3 in Molefe et al. (2005).

The main interest is testing the global null hypothesis of no treatment effect with respect to survival time, time to first tumor and number of tumors. (Molefe et al., 2005, also analyzed the detection time of tumors, but that data is not given here.) In case the global null hypothesis can be rejected, the deviations from the partial null hypotheses are of special interest.

Source

Molefe, D. F., Chen, J. J., Howard, P. C., Miller, B. J., Sambuco, C. P., Forbes, P. D. and Kodell, R. L. (2005). Tests for effects on tumor frequency and latency in multiple dosing photococarcinogenicity experiments. Journal of Statistical Planning and Inference 129(1–2), 39–58. doi:10.1016/j.jspi.2004.06.038

References

Hothorn, T., Hornik, K., van de Wiel, M. A. and Zeileis, A. (2006). A Lego system for conditional inference. The American Statistician 60(3), 257–263. doi:10.1198/000313006X118430

Examples

## Plotting data
op <- par(no.readonly = TRUE) # save current settings
layout(matrix(1:3, ncol = 3))
with(photocar, {
    plot(survfit(Surv(time, event) ~ group),
         lty =  1:3, xmax = 50, main = "Survival Time")
    legend("bottomleft", lty = 1:3, levels(group), bty = "n")
    plot(survfit(Surv(dmin, tumor) ~ group),
         lty = 1:3, xmax = 50, main = "Time to First Tumor")
    legend("bottomleft", lty = 1:3, levels(group), bty = "n")
    boxplot(ntumor ~ group, main = "Number of Tumors")
})
par(op) # reset

## Approximative multivariate (all three responses) test
it <- independence_test(Surv(time, event) + Surv(dmin, tumor) + ntumor ~ group,
                        data = photocar,
                        distribution = approximate(nresample = 10000))

## Global p-value
pvalue(it)

## Why was the global null hypothesis rejected?
statistic(it, type = "standardized")
pvalue(it, method = "single-step")

---

Class "PValue"

Description

Objects of class "PValue" represent the $p$-value, mid-$p$-value and $p$-value interval of the reference distribution.

Objects from the Class

Objects can be created by calls of the form

     new("PValue", \dots).

Slots

name:

Object of class "character". The name of the reference distribution.

p:

Object of class "function". The distribution function of the reference distribution.

pvalue:

Object of class "function". The $p$-value function of the reference distribution.

Methods

pvalue

signature(object = "PValue"): See the documentation for pvalue for details.

Note

Starting with coin version 1.3-0, this class is deprecated and will be replaced by class "NullDistribution". It will be made defunct and removed in a future release.

---

Computation of the $p$-Value, Mid-$p$-Value, $p$-Value Interval and Test Size

Description

Methods for computation of the $p$-value, mid-$p$-value, $p$-value interval and test size.

Usage

## S4 method for signature 'PValue'
pvalue(object, q, ...)
## S4 method for signature 'NullDistribution'
pvalue(object, q, ...)
## S4 method for signature 'ApproxNullDistribution'
pvalue(object, q, ...)
## S4 method for signature 'IndependenceTest'
pvalue(object, ...)
## S4 method for signature 'MaxTypeIndependenceTest'
pvalue(object, method = c("global", "single-step",
                          "step-down", "unadjusted"),
       distribution = c("joint", "marginal"),
       type = c("Bonferroni", "Sidak"), ...)

## S4 method for signature 'NullDistribution'
midpvalue(object, q, ...)
## S4 method for signature 'ApproxNullDistribution'
midpvalue(object, q, ...)
## S4 method for signature 'IndependenceTest'
midpvalue(object, ...)

## S4 method for signature 'NullDistribution'
pvalue_interval(object, q, ...)
## S4 method for signature 'IndependenceTest'
pvalue_interval(object, ...)

## S4 method for signature 'NullDistribution'
size(object, alpha, type = c("p-value", "mid-p-value"), ...)
## S4 method for signature 'IndependenceTest'
size(object, alpha, type = c("p-value", "mid-p-value"), ...)

Arguments

object	an object from which the $p$-value, mid-$p$-value, $p$-value interval or test size can be computed.
q	a numeric, the quantile for which the $p$-value, mid-$p$-value or $p$-value interval is computed.
method	a character, the method used for the $p$-value computation: either "global" (default), "single-step", "step-down" or "unadjusted".
distribution	a character, the distribution used for the computation of adjusted $p$-values: either "joint" (default) or "marginal".
type	pvalue(): a character, the type of $p$-value adjustment when the marginal distributions are used: either "Bonferroni" (default) or "Sidak". size(): a character, the type of rejection region used when computing the test size: either "p-value" (default) or "mid-p-value".
alpha	a numeric, the nominal significance level $\alpha$ at which the test size is computed.
...	further arguments (currently ignored).

Details

The methods pvalue, midpvalue, pvalue_interval and size compute the $p$-value, mid-$p$-value, $p$-value interval and test size, respectively.

For pvalue(), the global $p$-value (method = "global") is returned by default and is given with an associated 99% confidence interval when resampling is used to determine the null distribution (which for maximum statistics may be true even in the asymptotic case).

The familywise error rate (FWER) is always controlled under the global null hypothesis, i.e., in the weak sense, implying that the smallest adjusted $p$-value is valid without further assumptions. Control of the FWER under any partial configuration of the null hypotheses, i.e., in the strong sense, as is typically desired for multiple tests and comparisons, requires that the subset pivotality condition holds (Westfall and Young, 1993, pp. 42–43; Bretz, Hothorn and Westfall, 2011, pp. 136–137). In addition, for methods based on the joint distribution of the test statistics, failure of the joint exchangeability assumption (Westfall and Troendle, 2008; Bretz, Hothorn and Westfall, 2011, pp. 129–130) may cause excess Type I errors.

Assuming subset pivotality, single-step or free step-down adjusted $p$-values using max-$T$ procedures are obtained by setting method to "single-step" or "step-down", respectively. In both cases, the distribution argument specifies whether the adjustment is based on the joint distribution ("joint") or the marginal distributions ("marginal") of the test statistics. For procedures based on the marginal distributions, Bonferroni- or Šidák-type adjustment can be specified through the type argument by setting it to "Bonferroni" or "Sidak", respectively.

The $p$-value adjustment procedures based on the joint distribution of the test statistics fully utilizes distributional characteristics, such as discreteness and dependence structure, whereas procedures using the marginal distributions only incorporate discreteness. Hence, the joint distribution-based procedures are typically more powerful. Details regarding the single-step and free step-down procedures based on the joint distribution can be found in Westfall and Young (1993); in particular, this implementation uses Equation 2.8 with Algorithm 2.5 and 2.8, respectively. Westfall and Wolfinger (1997) provide details of the marginal distributions-based single-step and free step-down procedures. The generalization of Westfall and Wolfinger (1997) to arbitrary test statistics, as implemented here, is given by Westfall and Troendle (2008).

Unadjusted $p$-values are obtained using method = "unadjusted".

For midpvalue(), the global mid-$p$-value is given with an associated 99% mid-$p$ confidence interval when resampling is used to determine the null distribution. The two-sided mid-$p$-value is computed according to the minimum likelihood method (Hirji et al., 1991).

The $p$-value interval $(p_0, p_1]$ obtained by pvalue_interval() was proposed by Berger (2000, 2001), where the upper endpoint $p_1$ is the conventional $p$-value and the mid-point, i.e., $p_{0.5}$, is the mid-$p$-value. The lower endpoint $p_0$ is the smallest $p$-value attainable if no conservatism attributable to the discreteness of the null distribution is present. The length of the $p$-value interval is the null probability of the observed outcome and provides a data-dependent measure of conservatism that is completely independent of the nominal significance level.

For size(), the test size, i.e., the actual significance level, at the nominal significance level $\alpha$ is computed using either the rejection region corresponding to the $p$-value (type = "p-value", default) or the mid-$p$-value (type = "mid-p-value"). The test size is, in contrast to the $p$-value interval, a data-independent measure of conservatism that depends on the nominal significance level. A test size smaller or larger than the nominal significance level indicates that the test procedure is conservative or anti-conservative, respectively, at that particular nominal significance level. However, as pointed out by Berger (2001), even when the actual and nominal significance levels are identical, conservatism may still affect the $p$-value.

Value

The $p$-value, mid-$p$-value, $p$-value interval or test size computed from object. A numeric vector or matrix.

Note

The mid-$p$-value, $p$-value interval and test size of asymptotic permutation distributions or exact permutation distributions obtained by the split-up algorithm is reported as NA.

In versions of coin prior to 1.1-0, a min-$P$ procedure computing Šidák single-step adjusted $p$-values accounting for discreteness was available when specifying method = "discrete". This was made defunct in version 1.2-0 due to the introduction of a more general max-$T$ version of the same algorithm.

References

Berger, V. W. (2000). Pros and cons of permutation tests in clinical trials. Statistics in Medicine 19(10), 1319–1328. doi:10.1002/(SICI)1097-0258(20000530)19:10<1319::AID-SIM490>3.0.CO;2-0

Berger, V. W. (2001). The $p$-value interval as an inferential tool. The Statistician 50(1), 79–85. doi:10.1111/1467-9884.00262

Bretz, F., Hothorn, T. and Westfall, P. (2011). Multiple Comparisons Using R. Boca Raton: CRC Press.

Hirji, K. F., Tan, S.-J. and Elashoff, R. M. (1991). A quasi-exact test for comparing two binomial proportions. Statistics in Medicine 10(7), 1137–1153. doi:10.1002/sim.4780100713

Westfall, P. H. and Troendle, J. F. (2008). Multiple testing with minimal assumptions. Biometrical Journal 50(5), 745–755. doi:10.1002/bimj.200710456

Westfall, P. H. and Wolfinger, R. D. (1997). Multiple tests with discrete distributions. The American Statistician 51(1), 3–8. doi:10.1080/00031305.1997.10473577

Westfall, P. H. and Young, S. S. (1993). Resampling-Based Multiple Testing: Examples and Methods for $p$-Value Adjustment. New York: John Wiley & Sons.

Examples

## Two-sample problem
dta <- data.frame(
    y = rnorm(20),
    x = gl(2, 10)
)

## Exact Ansari-Bradley test
(at <- ansari_test(y ~ x, data = dta, distribution = "exact"))
pvalue(at)
midpvalue(at)
pvalue_interval(at)
size(at, alpha = 0.05)
size(at, alpha = 0.05, type = "mid-p-value")


## Bivariate two-sample problem
dta2 <- data.frame(
    y1 = rnorm(20) + rep(0:1, each = 10),
    y2 = rnorm(20),
    x = gl(2, 10)
)

## Approximative (Monte Carlo) bivariate Fisher-Pitman test
(it <- independence_test(y1 + y2 ~ x, data = dta2,
                         distribution = approximate(nresample = 10000)))

## Global p-value
pvalue(it)

## Joint distribution single-step p-values
pvalue(it, method = "single-step")

## Joint distribution step-down p-values
pvalue(it, method = "step-down")

## Sidak step-down p-values
pvalue(it, method = "step-down", distribution = "marginal", type = "Sidak")

## Unadjusted p-values
pvalue(it, method = "unadjusted")


## Length of YOY Gizzard Shad (Hollander and Wolfe, 1999, p. 200, Tab. 6.3)
yoy <- data.frame(
    length = c(46, 28, 46, 37, 32, 41, 42, 45, 38, 44,
               42, 60, 32, 42, 45, 58, 27, 51, 42, 52,
               38, 33, 26, 25, 28, 28, 26, 27, 27, 27,
               31, 30, 27, 29, 30, 25, 25, 24, 27, 30),
    site = gl(4, 10, labels = as.roman(1:4))
)

## Approximative (Monte Carlo) Fisher-Pitman test with contrasts
## Note: all pairwise comparisons
(it <- independence_test(length ~ site, data = yoy,
                         distribution = approximate(nresample = 10000),
                         xtrafo = mcp_trafo(site = "Tukey")))

## Joint distribution step-down p-values
pvalue(it, method = "step-down") # subset pivotality is violated

---