NEWS

rmsb 1.1-3

predict.blrm: fixed type='fitted' for binary Y; thanks Christian Hametner: <https://discourse.datamethods.org/t/bayesian-regression-modeling-strategies/6105/99>
removed Ocens function which is now in rms 7.0
changed to be compatible with storage of functions in rms::probabilityFamilies
updated to use Ocens2ord on top of Ocens, needs rms 8.0-0

blrm: added argument showopt to show Stan optimizer steps, default is FALSE
blrm: added options(rmsbmsg=FALSE) to allow the user to suppress information messages
lrmconppot.stan: fixed bug for dim(C2) p -> q
replaced use of survival::survConcordance and Hmisc::somers2 with concordancefit

Store cppo as deparsed for opt object from optimization (MLE) to prevent huge environment from being carried along
blrm: Non-downward compatible changes: removed keepsep argument
blrm: Added npcontrast argument for specification of priors for non-proportional-odds parameters in the constrained partial PO model through the use of contrasts, parallel to pcontrast for the PO part of the model; changed 3 stan code modules for this

blrm: added support for cmdstan and cmdstanr package using backend=
When using cmdstan uses bayesplot package for pairs plots
stanDx: new code for cmdstan diagnostics
stackMI: corrected erroneous implementation of file= where it was only applying to one of the imputations; also handled generally the conversion of functions in fit object to character strings for efficient serializing
Added non-exported function cluster so as to not depend on survival package
blrm: added sampling.args argument
blrm: trapped ppairs error in case plot region too small to hold graph
blrm: added pcontrast argument for user-specified priors on linear combinations of regression coefficients; required changing all Stan code
blrm: removed priorsd argument (NON-DOWNWARD-COMPATIBLE CHANGE)

blrm: got around bug in model.response not keeping class Ocens
soprobMarkovOrdPost: removed since Hmisc has soprobMarkovOrdm
lrmconppot.stan: removed redundant pi simplex (thanks Mark Jones <[email protected]>

Removed @title lines from Roxygen comments
Enhanced predict.blrm and changed prediction methods to use it
Changed storage of cppo function in fit object to character strings so as to not have saveRDS serialize it with all its environment data
Added lrmconppot.stan to implement flat and t-distribution priors for intercepts
blrm: added iprior argument for lrmconppot
predict.blrm: fixed fun= with finint=FALSE