Parametric Inference on Zero-Inflated Poisson distribution and its variants Santanu Chakraborty Mathematics Department UT Pan-American Abstract
The zero-inflated Poisson
distribution has found enough study importance
in the recent years for modeling count data, some biologoical phenomena
and also in software developments in computer science. This distribution
is derived from a usual Poisson probability distribution using a very
simple-minded approach. Starting with a Poisson random variable with
parameter $\lambda$, one can construct a Zero-Inflated Poisson (ZIP)
variable by reducing the mass at each of the non-zero values by a
constant proportion and increasing the mass at zero accordingly. This
was the original definition of a Zero-Inflated Poisson random variable.
Later there have been several generalizations of ZIP.
In this talk, we shall study some estimation and testing procedures for the Zero-inflated Poisson parameters and some of the interesting variants of ZIP. In particular, we demonstrate how to get rid of the nuisance parameters by conditioning on sufficient statistics. |