Zero-Inflated Count Regression
Zero-inflated count regression models may be used to predict a discrete non-negative integer variable with excess zeroes. The models contain two components: one for the underlying counts and a second for extra structural zeroes. Statgraphics 19 contains a procedure for fitting both standard zero-inflated count regressions, often based on a Poisson distribution, and hurdle models that use a truncated distribution for the positive counts and a second distribution for all of the zeroes.