NEWS

erglm 0.2.0

erglm_scm_forward()/erglm_scm_backward() gain a criterion argument, supporting "aic"/"bic"-based term selection in addition to the existing "p-value" default. The SCM history (erglm_scm_history()) gains a criterion column recording which selection rule was applied in each forward/backward step (#7).
The bundled erglm_data dataset's continuous columns (aucss, cmaxss, biomarker_change, ae_duration) are now rounded to two decimal places, a more realistic degree of precision; underlying values are otherwise unchanged.
erglm no longer depends on rlang, withr, tibble, or dplyr (mvtnorm remains its only runtime dependency besides base stats). As a result, erglm_predict(), simulate.erglm_model(), and the bundled erglm_data dataset now return/are plain data frames rather than tibbles, and erglm_predict()'s fit_link/se_link/fit_resp columns lose the observation-index names they used to carry as a predict.glm() artifact; underlying values are otherwise unchanged. tibble has also been dropped from Suggests entirely, since the vignettes now build their example data with plain data.frame() calls too.
Fixed several documentation gaps across help pages, vignettes, and the README: missing descriptions, undocumented argument defaults, and broken cross-references, including several references to erplots' er_vpc_plot(), which has been replaced by er_vpc_add_simulated().

CRAN resubmission, addressing reviewer feedback on the 0.1.0 submission:

Self-references to 'erglm' in the DESCRIPTION Description field are now single-quoted, per CRAN's software-name convention.
DESCRIPTION now declares Additional_repositories: https://djnavarro.r-universe.dev, the repository from which the optional Suggests dependency 'erplots' (not on CRAN) can be installed, per CRAN policy on declaring where to obtain such packages.
DESCRIPTION's Description field now capitalises Poisson and Gaussian (they name the eponymous distributions/families), per further reviewer feedback; binomial and gamma are left lowercase, as neither is an eponym.

Initial CRAN release.

Model fitting and prediction for exposure-response models based on glm() (erglm_model(), erglm_predict()), supporting arbitrary glm() 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-term erglm_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.