Table 1 Possible statistical models and assumptions for count data
Model | Assumptions |
|---|---|
Poisson regression | Mean and variance are equal |
Negative binomial regression | Overdispersion in the data, where variance is greater than mean |
Zero-inflated model | Two sources of zeros: “structural” and “sampling”. Structural zeros are due to some specific structure in the data and sampling zeros are due to Poisson or negative binomial distribution |
Hurdle model | All zeros are from a “structural” source and all positive values are from a “sampling source” |