B1108
Title: Count data regression with excess zeros: A flexible framework using the GLM toolbox
Authors: Christian Kleiber - Universitaet Basel (Switzerland)
Achim Zeileis - Universitaet Innsbruck (Austria) [presenting]
Abstract: The hurdle model is a two-part model for count data with extra zeros, comprising a binary response part for zeros vs. non-zeros and a zero-truncated count distribution for the positive counts. We show how not only the binary part but also the count component can be analyzed within the GLM framework. This paves the way for flexible extensions of the hurdle model using methods from the extended GLM toolbox such as additive nonlinear terms, boosting, bias reduction, etc. Similarly, it is straightforward to apply visualization techniques such as rootograms, effect displays, or residual plots.
