A generalized linear model lets you choose the family and link yourself — Gaussian, binomial, Poisson, Gamma, inverse Gaussian, quasi-binomial or quasi-Poisson, each with its standard links.
Generalized linear models extend OLS to non-normal responses by combining a link function (mapping the linear predictor to the conditional mean) with a distribution family (specifying the variance). The specialised entries (linear_regression, logistic_regression, poisson_regression, negbin_regression) cover the four most common combinations; glm_general is for everything else.
Common less-common combinations: Gamma + log link for positive continuous data with multiplicative effects (insurance losses, hospital stays); inverse Gaussian + 1/μ² link for first-passage times; binomial + probit link for bioassay; binomial + cloglog for grouped survival data. Quasi-binomial and quasi-poisson allow dispersion ≠ 1 with the same link.
Response + predictors + intercept toggle + family + link.
β + SE + Wald z + p + 95% CI per coefficient + null + residual deviance + LRT χ² vs intercept-only + AIC + dispersion + exp(β) for log / logit links.
Each family has its own appropriate effect-size reporting (rate ratio for log link, odds ratio for logit, multiplicative effect for Gamma log link, etc.).