"Identification and Accommodation of Overdispersion in Discrete Response Regression Models"
April 1, 2008 @ 3:30 pm - 4:30 pm
Joseph Hilbe, Arizona State University
Location: BRB - II/III - Room 253
TITLE: Identification and Accommodation of Overdispersion in Discrete Response
ABSTRACT: Overdispersion is a perennial problem for binomial and count response regression modeling. We discuss the nature and scope of overdispersion, how to define and determine apparent from real overdispersion, and offer methods of dealing with both. In particular we look at how overdispersion affects binomial or proportional logistic regression, and count models such as Poisson, generalized Poisson, and the many varieties of negative binomial; e.g. NB-2, NB-1, NB-C, NB-H, and NB-P models.