Count & Rate Data

Summary

Develops count regression starting from the Poisson distribution's mean-equals-variance constraint and the log link that produces incidence rate ratios on exponentiation, illustrated by a mastitis-in-dairy-herds example. Diagnoses overdispersion when residual variance exceeds the mean and introduces negative binomial regression with its additional dispersion parameter, then walks through zero-inflated and hurdle models for excess zeros, plus offsets that handle person-time at risk and let the model produce true rate ratios rather than count ratios.