Changelog
Source:NEWS.md
assessor 1.3.1
CRAN release: 2026-04-20
-
dpit()now rejects non-TobitVGAM::vglmobjects and non-constant-scale VGAM Tobit fits with informative errors. -
dpit_2pm()now requires exactly one model or supplied-probability input for each component and uses one common calculation path. - Corrected the
ord_curve()axis documentation and threshold label, and clarified the plotting documentation forord_curve()andquasi_plot(). -
dpit_ordi()now respects the ordering supplied inleveland validates the fitted category-probability matrix. - Corrected the zero-inflated negative-binomial bandwidth calculation to use the negative-binomial CDF.
- Goodness-of-fit bootstrap refits now preserve offsets for binomial, Poisson, and negative-binomial regression models.
-
gof_disc()now requiresBto be a positive integer. - Fixed an error where
scale = "uniform"did not produce a uniformly scaled QQ plot. - Fixed the negative-binomial goodness-of-fit bootstrap so that each replicate is refitted using the simulated response.
-
dpit(),dpit_2pm(), and the distribution-specific calculators for binary, Poisson, negative binomial, ordinal, zero-inflated, Tobit, and Tweedie outcomes now returndpitobjects. Useresiduals()to extract values,summary()to summarize them, andplot()to draw QQ plots. Thescaleargument is now supplied toresiduals(),summary(), orplot(), and graphical arguments are supplied toplot(). - The
print()andsummary()methods now label the number of residuals as the sample size.
assessor 1.3.0
CRAN release: 2026-03-22
Breaking changes
- Renamed all functions with prefix
resid_*todpit_*. - Unified
resid_disc(),resid_zeroinfl(), andresid_semiconti()into a single S3 generic functiondpit(). - Renamed
resid_quasi()toquasi_plot().
New features
- Added
dpit_2pm()(formerlyresid_2pm()). - Introduced direct DPIT computation functions:
- Added
gof_disc(), an S3-based goodness-of-fit testing function for discrete outcome regression models. - Updated
quasi_plot()to an S3-based interface.
assessor 1.2.0
- Add
bballHRdata set which is used as an example for assessing a discrete outcome regression.