Summary

Distribution	Useful for…	Canonical link function	Interpretation
Gamma	Right skewed distributions	log	Percent change in outcome associated with unit change in predictor.
Poisson	Count outcomes with low dispersion	log	Percent change in outcome associated with unit change in predictor.
Negative-Binomial	Count outcomes with high dispersion	log	Percent change in outcome associated with unit change in predictor.
Zero-inflated	Any outcome with excess of zeros generated by a secondary process	Logit for zero-inflation component	Change in odds of outcome associated with unit change of the predictor
Logistic models	Categorical outcomes with two levels	logit	Change in odds of outcome associated with unit change of the predictor
Ordinal outcomes	Categorical outcomes with ordered levels (incl. Likert scales)	logit	Change in odds of belonging to higher levels associated with unit change of the predictor