Zero-inflated model

A zero-inflated model is a statistical model built on a zero-inflated probability distribution, meaning a distribution that allows for a higher frequency of zero valued observations than standard count distributions predict. Such models are used to analyze count data, such as emergency room visits, fish caught, or insurance claims, in cases where the observed number of zeros is greater than a Poisson or negative binomial distribution would expect. Data with this pattern are described as zero inflated, and using an ordinary count model on them tends to underestimate how often a zero outcome actually occurs.