Far East Journal of Theoretical Statistics
Volume 21, Issue 2, Pages 203 - 216
(March 2007)
|
|
MULTI-CLASS MODELS FOR CORRELATED ZERO-INFLATED COUNT DATA
Getachew A. Dagne (U. S. A.)
|
Abstract: In many areas of research, count
data may exhibit a significantly higher percentage of zeros. In this case,
standard generalized linear models (GLM) may not fit the data well. Methods
suggested in the literature to incorporate excess zeros and fit the data
relatively better than the standard GLM are zero-inflated Poisson model,
left-censored model or two-part models. We extend these models to more general
models that accommodate more than two parts. This article presents
three-component growth mixture models which incorporate both overdispersion and
excess zeros. The components may be resultants of some intervention (treatment)
that changes the rare event generating process. The models are also expanded to
take into account the heterogeneity and correlations existing in the data. The
proposed methods are applied to count data on the assessment of an efficacy of
pesticides in controlling the reproduction of white flies. |
Keywords and phrases: random effects, EM algorithm, zero-inflated, Poisson
regression, growth mixture model. |
|
Number of Downloads: 272 | Number of Views: 722 |
|