Changelog
Source:NEWS.md
erglm 0.1.1
CRAN release: 2026-08-08
CRAN resubmission, addressing reviewer feedback on the 0.1.0 submission:
- Self-references to ‘erglm’ in the
DESCRIPTIONDescriptionfield are now single-quoted, per CRAN’s software-name convention. -
DESCRIPTIONnow declaresAdditional_repositories: https://djnavarro.r-universe.dev, the repository from which the optionalSuggestsdependency ‘erplots’ (not on CRAN) can be installed, per CRAN policy on declaring where to obtain such packages. -
DESCRIPTION’sDescriptionfield now capitalisesPoissonandGaussian(they name the eponymous distributions/families), per further reviewer feedback;binomialandgammaare left lowercase, as neither is an eponym.
erglm 0.1.0
Initial CRAN release.
- Model fitting and prediction for exposure-response models based on
glm()(erglm_model(),erglm_predict()), supporting arbitraryglm()families; binomial, poisson, gaussian, and gamma are tested and officially supported end to end (fitting, prediction, SCM significance testing, and simulation). - Stepwise covariate modelling (
erglm_scm_forward(),erglm_scm_backward(),erglm_scm_history()), built on the single-termerglm_add_term()/erglm_remove_term()helpers. - Simulation support (
erglm_fun(),simulate.erglm_model()) for drawing replicate responses from a fitted model, with sampled coefficients and both expected and simulated response columns. - Interoperability with the companion erplots package via
er_predict()/er_simulate()/er_summary()methods, registered lazily so erglm has no hard dependency on erplots or on any plotting package. -
erglm_link()/erglm_invlink(), discoverable wrappers around a fitted model’s link and inverse-link functions. - An example dataset,
erglm_data, with binary, count, and continuous/right-skewed continuous response columns for demonstrating each supported family.