Counts have variance far above their mean. Why consider negative binomial instead of Poisson?
StatisticsLevel: IntermediateDiscrete probabilistic models
Counts have variance far above their mean. Why consider negative binomial instead of Poisson?
Hints
- Poisson has a mean–variance constraint.
- A second parameter adds flexibility.
Answer
Negative binomial allows overdispersion through an extra parameter. Poisson ties variance to mean; also investigate heterogeneity, dependence and excess zeros.
Source and licence: SciPy — discrete probability distributions · BSD-3-Clause · answer: Quantoros editorial