Changes in version 1.8-0 (2026-04-14) Features o Implement residuals.mmlt (allowing sandwich estimation and permutation tests) and update.mmlt (a little bit faster refitting). o Remove type = "Lambdapar" from coef.mmlt, use type = "Lambda". Changes in version 1.7-4 (2026-03-06) Features o Add type = "Omega" in coef. type = "Lambda" now always returns a unit lower-triagular matrix. This now matches the notation used in papers. o Generate references from REFERENCES.bib. Bugfixes o Determine constant columns in design matrices correctly. Changes in version 1.7-3 (2025-12-12) Bugfixes o Remove constant columns only when dofit is TRUE. Changes in version 1.7-2 (2025-11-12) Bugfixes o R() was handled incorrectly. Spotted by Lukas Graz. Changes in version 1.7-1 (2025-10-31) Bugfixes o Handle infinite values when computing quantiles from response objects. Changes in version 1.7-0 (2025-10-19) Features o mltoptim() features nlminb for unconstrained optimisation. o mltoptim() features optim for unconstrained optimisation. o mltoptim() defaults changed for all algorithms. o mltoptim() refrains from generating random starting values now. Consequently, results are fully reproducible in the absence of user-defined seeds. o The log-likelihood for infeasible parameter configurations is now always -Inf. o Argument scale was renamed to scaleparm because of a conflict in tram::Survreg. o Compute better starting values, especially for interval-censored outcomes using the fast Turnbull estimators provided by icenReg. o Implement (internal) pre-transformation based on nonparametric unconditional distribution to be used by tram. Bugfixes o summary.mmlt works now. o Improved error messages in mmlt. o R(, as.R.interval = TRUE) produces correct empirical likelihood for numeric outcomes. o mmlt only uses optimisers able to approximate the Hessian by default. o simulate.ctm might have got internal dimensions wrong. Changes in version 1.6-6 (2025-06-17) Features o Simpler switch for computing numerically approximated Hessians in mltoptim. Changes in version 1.6-5 (2025-05-09) Bugfixes o rep to rep_len when possible. o Avoid partial argument matches. Changes in version 1.6-4 (2025-04-29) Features o Improve R() for interval censored data. o Add constrOptim to mltoptim. Bugfixes o Handle fixed starting values in mmlt correctly. Changes in version 1.6-3 (2025-03-20) Features o Add as.Surv method for numeric vectors. Changes in version 1.6-2 (2025-01-29) Bugfixes o Fall back to coneproj if quadprog fails to get starting values. o Stabilise bugfixes and avoid ATLAS issues. Changes in version 1.6-1 (2024-11-18) Bugfixes o Try harder to obtain starting values meeting constraints, issue a warning if this failed. Changes in version 1.6-0 (2024-09-23) Bugfixes o Allow interval censoring using -Inf and Inf. Starting values are affected but the corresponding observations contribute zero to the log-likelihood. o Make sure R() does not drop any unused levels (unlike factor()). New features o mmlt moved from tram allowing arbitrary mixes of continuous and discrete data, including exact and censored observations for the same variable and thus also allowing non-informative missings encoded as -Inf, Inf, see below. Corresponding logLik and estfun methods feature newdata and weights arguments, as the methods for mlt always did. o R() translates missing response values to -Inf and Inf interval censoring. This means that missing values are allowed in response variables and the corresponding observations do not contribute to the log-likelihood. In contrast to na.omit, the (technical) dimension of the training data is not changed. Changes in version 1.5-2 (2024-08-24) Bugfixes o Handle Surv objects with smaller than zero values in R.Surv(..., as.R.interval = TRUE). New features o Add maxeval = 1000L for nloptr. o Report nloptr more detailed status messages. Changes in version 1.5-1 (2024-04-16) Bugfixes o Remove warning when dealing with binary factors. Changes in version 1.5-0 (2023-12-14) Bugfixes o Do no use zero as lower bound (when running R(Surv(...), as.R.interval = TRUE)) except when zero is first death time. New features o Add slots to mmlt for externally computing transformations and derivatives thereof, needed by tram::mmlt. Changes in version 1.4-9 (2023-08-21) Bugfixes o Make sure R(Surv(...), as.R.ordered = TRUE) matches Kaplan-Meier for right-censored and possibly tied observations. Such responses give NPML estimates. New features o R(Surv(...), as.R.numeric = TRUE) can be used to fit smooth models by maximising the nonparametric likelihood. Changes in version 1.4-8 (2023-06-30) New features o Better support for frailty parameter. Changes in version 1.4-7 (2023-05-22) Bugfixes o Starting values might end up being NA; set to zero. New features o mlt(..., theta = theta, dofit = TRUE) returns a "fitted" mlt model with coefficients theta. If theta is missing, an unfitted model is returned. Changes in version 1.4-6 (2023-04-14) Bugfixes o Register internal methods. New features o Re-add support for nloptr in mltoptim(). Changes in version 1.4-5 (2023-03-09) Bugfixes o Fix S3 argument mismatches. o Try to detect negative standard errors in case optimisation failed silently. Changes in version 1.4-4 New features o Improve numerical stability in simulate.ctm. Changes in version 1.4-3 (2022-10-31) New features o predict and plot have new types involving logs. Bugfixes o plot() ignored lwd. Changes in version 1.4-2 (2022-06-30) New features o as.R.interval allows evaluation of nonparametric likelihood for smoothly defined models. o New Laplace link function, contributed by Ainesh Sewak. o New cauchit link function. Bugfixes o drop = TRUE in score for shift-scale models. o Eliminate fallout of fix of PR#17616 in Rout.save files. Changes in version 1.4-1 (2022-03-31) New features o update allows fixing parameters. Bugfixes o Fix discrete gradient for shift-scale models. Changes in version 1.4-0 (2022-01-14) New features o Add infrastructure for shift-scale transformation models. Documentation o Gradient and estfun always returned negative scores, this is documented now. Bugfixes o Arguments passed to mlt via dots were silently ignored. o Remove nloptr dependency for the time being. Changes in version 1.3-2 (2021-09-02) New features o Add as.double for response objects (replaces trtf:::.R2vec). Changes in version 1.3-1 New features o Improved numerical stability for censored data. Changes in version 1.3-0 (2021-03-09) New features o New argument as.R.ordered allowing numeric and survival responses to be coded as ordered factors, for nonparametric maximum likelihood estimation. o Allow sparse model matrices. This is useful for nonparametric maximum likelihood estimation with many distinct outcomes. o Some speed-ups. Bugfixes o Plotting of quantiles sometimes failed because inversion of cdf was not possible for certain quantiles. These are now removed before plotting. o Fitting models to interval censored responses containing intervals c(-Inf, Inf) failed. o Always return names score matrices and residuals. Changes in version 1.2-3 (2021-02-24) New features o Improve documentation. Changes in version 1.2-2 (2021-02-12) Bugfixes o Sampling from unconditional models did not pay attention to number of observations. Changes in version 1.2-1 (2021-02-05) New features o Quantiles and thus simulations are now computationally more exact and more robust. The unnecessary interpolate argument to predict and simulate is now ignored. o Adjust contrasts a fixed parameter contributes to. o Return numerically determined Hessians upon request. o Implement frailty error distributions, experimentally and internal only. o Implement cure mixture models, experimentally and internal only. o Improve computations of log-probabilities. Bugfixes o Discrete hazard functions were incorrect. Changes in version 1.2-0 (2020-05-12) o Add exponential distribution (for Aalen additive hazards models). o Pay attention to model class when computing cumulative hazards. o Add log-cumulative hazards, log-odds, and odds for predictions and plots. Changes in version 1.1-2 (2020-02-14) o Allow permutations of single variables. o Update citation info. Changes in version 1.1-1 (2020-01-10) o Try harder to invert Hessians. o Update reference output. Changes in version 1.1-0 o Add support for nloptr (still experimental and thus switched off by default). o Make sure coef() always returns named argument. o Fix problem in as.Surv reported by Balint Tamasi. Changes in version 1.0-7 (2019-10-06) o Less paranoia in bugfixes.R. Changes in version 1.0-6 (2019-10-04) o Return Hessian for fixed parameters if requested. o Fix subsetting problem in R.numeric. o Allow to update offsets. Changes in version 1.0-5 (2019-06-21) o Add a bread method. o Check response variable against observations in data. o Make sure integers larger zero are handled correctly in R. o Implement resid method, ie the score wrt a constant. o Cox examples with Bernstein polynomials of log-time. Changes in version 1.0-4 (2018-12-03) o Arguments K and cheat where ignored by confband when newdata had multiple rows. o Computation of starting values more robust now. o Order of fixed parameters (fixed argument to mlt) might have been wrong due to incomplete matching. Changes in version 1.0-3 (2018-09-17) o Add lty argument to plot.ctm. o update needs free coefficients only. o Internal interface changes. Changes in version 1.0-2 (2018-06-22) o Make sure transformation functions outside bounds are minus or plus Inf. o Initial guestimates for ordered responses were incorrect and may potentially have led to nonsense results. o Some smaller improvements in computation of log-likelihoods and scores with respect to accuracy and speed. o print respects options(digits). Changes in version 1.0-1 (2018-04-20) o estfun, parm = coef(object, fixed = TRUE)) evaluates scores for all model parameters, including fixed ones. o logLik(..., newdata, w) ignored weights w when newdata was given. Same problem was also fixed for estfun. Changes in version 1.0-0 (2018-03-08) o A paper describing version 1.0-0 of the mlt, basefun, and variables packages was accepted for publication in the Journal of Statistical Software 2018-03-05. o Documentation updates. Changes in version 0.2-2 (2018-02-23) o Use coneprog for getting the starting values. o logLik and estfun accept matrices as parm argument for the evalution of log-likelihoods and scores with subject-specific parameters (for example in transformation trees or forests and boosting procedures. Changes in version 0.2-1 (2017-12-12) o q is forwarded to qmlt by predict.ctm now. o p is now prob in qmlt and thus predict.ctm. o Update citation. Changes in version 0.2-0 (2017-06-20) o Most Likely Transformations will be published in the Scandinavian Journal of Statistics. o Import package alabama. o as.Surv(R(Surv(...))) returns Surv(...), useful for converting output by simulate to Surv objects. Changes in version 0.1-3 (2017-04-05) o Add subset argument to update (for faster transformation trees and forests). o Sum over score contributions with positive weight only when evaluating the gradient. o Having all response observations being interval-censored is allowed again (too heavy checking was in place). o Don't try to numerically check KKT conditions automatically. o Check for unused arguments in dots where necessary. o Make sure the score doesn't get too large (avoid division by near zero probabilities). o Improve survfit to compute non-parametric unconditional probabilities for obtaining starting values in the presence of censoring and truncation. Changes in version 0.1-2 (2017-02-14) o logLik with newdata argument ignored parm and weights arguments. o estfun now also has a newdata argument. o Correct axes labelling when plotting quantile functions. o make sure names are correct in coef(model, fixed = FALSE). o check if any exact or interval-censored response with non-zero weight exists before trying to fit the model. o make checks a little more robust against huge diffs under Windows. Changes in version 0.1-1 (2016-11-29) o Fix two bugs in computation of log-likelihood for possibly left-truncated responses such as Surv(start, time, status). Changes in version 0.1-0 (2016-11-10) o Add augmented lagrangian minimization (auglag() from package alabama). o Make optimiation procedure more general and adaptive, allow users to change defaults and even add their own optimiser. Changes in version 0.0-34 (2016-09-13) o fix bug when calling survfit for computing initial probabilities. o add bysim argument to simulate. o make sure checkGrad is respected by update. o predict computes q with K elements if not given (as plot always did). Changes in version 0.0-33 (2016-05-25) o Make sure times are ordered before calling survival::summary.survfit. Changes in version 0.0-32 (2016-05-25) o Introduce as.mlt generic. o Introduce a coef slot in ctm objects and a corresponding coef<- and coef method for setting and extracting coefficients to and from unfitted conditional transformation models. o predict, simulate and plot work on ctm objects (with meaningful coefficients) now. Changes in version 0.0-31 o Some small improvements wrt run time and memory consumption. o Use theta = coef(object) as default starting parameters in update(). o logLik has a new newdata argument. o simulate has a new q argument. Changes in version 0.0-30 (2016-02-20) o The mlt package was first published on CRAN.