- interaction terms between latent and / or manifest predictors can now be included in the model
- added checks for OpenMP to avoid compilation errors on MacOS
- tests are skipped on CRAN, as minor numerical deviations cannot be avoided
- refactoring of the whole codebase
- major change to the C++ files; likelihood estimation is now based on RcppArmadillo and matrix algebra
- major changes to the R files; partable, modellist and so on were adapted to the matrix-based computations
- codebase now supports interaction terms, but this feature will be released with the next minor version (not tested enough right now)
- minor changes to the C++ files due to upcoming changes in Rcpp package (STRICT_R_HEADERS)
- added some details in the documentation of the main function
- skipping two tests failing on MacOS (i.e. applications in which only one latent covariate is considered)
- lowered numerical precision for test on Poisson regression with three manifest covariates (for Solaris)
- added a more detailed summary-function
- Added a
NEWS.md
file to track changes to the package.