Winkelmann
Fakultäten » Wirtschaftswissenschaftliche Fakultät » Volkswirtschaftslehre, Institut für » Prof. Dr. Rainer Winkelmann » Winkelmann
Completed research project
| Title / Titel | Robust estimation of zero-inflated count data models | ||||
|---|---|---|---|---|---|
| Abstract (PDF, 14 KB) | |||||
| Summary / Zusammenfassung | Count data often display too many, or extra, zeros, invalidating the assumption of a Poisson or negative binomial distribution. In response, modified count data models have been suggested, including the zero-inflated Poisson and the zero-inflated negative binomial model, and these are increasingly used in practice. While the Poisson distribution (and the negative binomial distribution for given dispesion parameter) is a linear exponential family, zero-inflated distributions are not. This has consequences for the behavior of maximum likelihood estimators under misspecification. While the Poisson model estimates regression parameters consistently under very general assumptions, zero inflated models lead to efficient estimators if correctly specified, but are inconsistent otherwise. We explore the trade-off between efficient and robust estimation in an extended Monte Carlo study. | ||||
| Project leadership and contacts / Projektleitung und Kontakte |
|
||||
| Funding source(s) / Unterstützt durch |
Universität Zürich (position pursuing an academic career) |
||||
| Duration of Project / Projektdauer | Jul 2008 to Jun 2010 |