- Bugfix: some arguments passed as variables were not found. This was introduced by moving away from NSE and not exporting some internal functions.
- Bugfix: error with printing models called with
`tramME::`

. - Documentation: Clarify that the print method of
`VarCorr.tramME`

shows the random effect standard deviations by default. - New functionality:
`coef.tramME`

now has a`complete`

argument that returns all coefficients (baseline, fixed, random) of the model as a named vector - New functionality: Access variances and covariances of penalized
parameters (coefficients of smooth terms and random effects) through the
`parm`

argument of`vcov.tramME`

. - New functionality/change of behavior: The new
`ok_warnings`

argument of`optim_control`

provides control over what warning messages are reported during optimization.`summary.tramME`

doesn’t report the number of warnings during optimization anymore. - New functionality: New scales are available in
`predict.tramME`

and`plot.tramME`

. - New functionality:
`confband.tramME`

calculates confidence bands and pointwise confidence intervals for various scales of the conditional outcome distribution. - Documentation: Update documentation of the model functions with additional details, examples and references.
- Processing random effects formula elements with
`reformulas`

.

- Refactor code, clean up documentation, unexport internal functions.
- Improve numerical robustness (log-likelihood, edf).

- Activate the multi-core capabilities of
`TMB`

`Resp()`

function to extend the functionality of`Surv`

objects`trafo`

option for`plot.smooth.tramME`

to plot monotonic transformations of the smooth terms- Faster and more robust
`vcov`

and`edf_smooth`

calculations. `simulate`

method (with limited functionality in this version) returned- Updated additive
`tramME`

package vignette

- Adapted matrix type conversions to Matrix 1.5.0.
- Fixed typo in vignette
`mixed-effects-additive-models`

.

- Updated vignette
`mixed-effects-additive-models`

. - Cleaned up the documentation.

- Updated test utilities to avoid issues with MKL on CRAN.

- Smooth shift terms as defined by
`mgcv`

. Currently only`s()`

smooths are allowed. - Cleaned up the source code and the documentation.
- The official reference is Tamasi and Hothorn (2021), vignette was also updated.
- New vignette with examples of mixed-effects transformation models with smooth shift terms.
- Some methods have been moved to development branch for proper
testing and refactoring (
`Aareg`

,`lpterms`

,`simulate`

,`parboot`

,`trafo`

). These features may eventually find their way back to the package, probably in a slightly changed form.

- Fixed
`duplicate.tramTMB`

method to work with TMB 1.8.0.

- Added R-Forge URL

- Fixed uninitialized value problem indicated by valgrind

- Updated internal functions for estimation transformation models with
TMB:
- Fixed effects only models can be estimated
- Exported functions (
`coef`

,`logLik`

) communicate with the TMB model more smoothly - Updated internal structure to help future development (not exported currently)

- Fixing coefficients with the argument
`fixed = c(name = value)`

- New model classes in
`SurvregME`

using the parameter fixing option `AaregME`

extends the (parametric) Aalen regression model of`tram`

with mixed-effects.- Out-of-sample log-likelihoods using the
`newdata`

argument of`logLik`

- Score residuals with the
`resid`

method. When frequently recalculated (e.g. in boosting), setting`resid = TRUE`

in model definition increases efficiency. - Updating models via the
`update`

method - Setting observation weights and offsets efficiently (activated with
`do_update = TRUE`

). Currently, only possible through directly manipulating the`tmb_obj`

of the model. - Parametric bootstrap with the
`parboot`

method - New optimization options (e.g. internal scaling of fixed effects
design matrix to improve convergence, more sensible initial values) and
improved control over the optimization process with
`optim_control()`

- Various methods (including
`analytical`

in the case of fixed effects only models) to calculate the Hessian; trying harder to invert the Hessian in numerically unstable cases. - Calculate the linear predictor with
`type = "lp"`

in`predict`

- Several additional methods to help the user working with
`tramME`

objects:`model.frame`

,`model.matrix`

,`fitmod`

,`duplicate`

. - Improved unit testing
- Improved documentation
- Demo for IPD meta-analysis

- fixed bug in setting error distributions of ‘dummy’ ctms for predict and simulate methods
- updated Figure 6 in vignette, because the bug above affected predict in the case of CorlME
- fixed bug in unit test for simulate that caused error with mlt 1.2-1
- fixed simulate output structure with
`what = "joint"`

option

- Added author ORCID
- Fixed CRAN issue with unit test using MKL
- Figure 3 color/legend changed in Vignette

- Fixed numerical precision problem in unit tests

- First CRAN version