Winai Bodhisuwan, Siriporn Samutwachirawong, and Prasit Payakkapong

DOI: 10.14456/sjst-psu.2018.174

pp. 1428 - 1436

Abstract

In this study, we propose the zero-inflated negative binomial-Erlang distribution for modeling the number of highly
pathogenic avian influenza H5N1 outbreaks that were reported at the subdistrict level in Thailand. We provide the probability
mass function, mean, variance, skewness, and kurtosis for the zero-inflated negative binomial-Erlang distribution. In addition,
this distribution provides a better fit compared to the zero-inflated Poisson and the conventional zero-inflated negative binomial
distributions. In Thailand, a highly pathogenic avian influenza virus of the H5N1 subtype was first confirmed in humans in 2004.
The epidemic has only caused sporadic outbreaks in Thailand during the second epidemic wave, particularly in the central and
northern regions. The zero-inflated Poisson distribution can be used efficiently in epidemiologic count data with an excess
number of zeros. However, when the data exhibit overdispersion, an alternative to the zero-inflated Poisson distribution may be
considered such as the zero-inflated negative binomial-Erlang distribution.