The Zero-Inflated Poisson (ZIP) distribution, typically assumed for modeling count data with excess of zeros, assumes that with probability p the only possible observation is zero, and with probability 1 p a Poisson(y) random variable is observed. Both the probability p and the mean y may depend on covariates. In this paper we discuss and apply Bayesian inference based on matching priors and on higher-order asymptotics to perform accurate inference on y only, even for small sample sizes.
Modern Bayesian inference in Zero-Inflated Poisson models
RULI, ERLIS;VENTURA, LAURA
2012
Abstract
The Zero-Inflated Poisson (ZIP) distribution, typically assumed for modeling count data with excess of zeros, assumes that with probability p the only possible observation is zero, and with probability 1 p a Poisson(y) random variable is observed. Both the probability p and the mean y may depend on covariates. In this paper we discuss and apply Bayesian inference based on matching priors and on higher-order asymptotics to perform accurate inference on y only, even for small sample sizes.
File in questo prodotto:
Non ci sono file associati a questo prodotto.
Pubblicazioni consigliate
I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.
