R package glmdr: Generalized Linear Models Done Right. The name stands for ”generalized linear models done right”, where ”done right” means it correctly handles cases where the maximum likelihood estimate (MLE) does not exist in the conventional sense. Only does discrete generalized linear models and only those that are exponential family (because only exponential families have good theory about existence of MLE). Also does log-linear models for contingency tables and multinomial logistic regression, which it handles as conditional distributions of Poisson regression. It provides valid hypothesis tests and confidence intervals even when the MLE are ”at infinity” in terms of canonical parameters or ”on the boundary” in terms of mean value parameters, following