NEWS

nhm 0.1.2 (2025-08-31)

Constrained optimization now possible via nlminb.
Censoring of non-final observations possible for misclassification models.
Left truncation possible for misclassification models.
Introduced informative observation testing via inform option.
The control argument can now be a named list rather than a nhm.control object.
Phase-type semi-Markov models can be accommodated for "bespoke" type models using the phasemap option.
Parallel computing also now used for performing the forward algorithm for misclassification models and to calculate the Hessian using finite differences .
Links to functions in other packages fixed within help documentation.

Fixed bug in model.nhm that occurred if a bespoke model was specified without including the number of parameters as an attribute in the intens function.
Fixed bug in ematrix.nhm when a null covvalue is supplied.
Fixed an error in the way in which individual component-wise p-values are calculated for the score test (in print.nhm_score). The vignette has also been updated to give correct results for the example.
Fixed bug when a model of Weibull type is specified with no time-dependent terms.
Fixed bug in dataprocess.nhm that occurred if a model with covariates and estimated initp was specified with firstobs="misc" (thanks to Emmett Kendall for the report).