|
Activity Number:
|
270
|
|
Type:
|
Contributed
|
|
Date/Time:
|
Tuesday, August 4, 2009 : 8:30 AM to 10:20 AM
|
|
Sponsor:
|
Section on Statistical Computing
|
| Abstract - #305555 |
|
Title:
|
The Model Selection of Zero-Inflated Mixtured Poisson Regression
|
|
Author(s):
|
Huaiye Zhang*+ and Inyoung Kim
|
|
Companies:
|
Virginia Polytechnic Institute and State University and Virginia Polytechnic Institute and State University
|
|
Address:
|
Dept. of Statistics, Blacksburg, VA, 24061,
|
|
Keywords:
|
Bootstrap ; Expectation Maximization algorithm ; Mixing Component ; Zero-Inflated Poisson regression ; Score test
|
|
Abstract:
|
Poisson regression provides a standard framework for the analysis of the counting data. However in many application, count data has many zeros and also has the mixture distributions. The zero-inflated mixture Poisson regression can be handled for the data. However, it is not obvious to choose a zero-inflated Poisson model without any statistical evidence and also difficult to select the number of mixing components in mixture distributions. Hence, in this paper, we propose a score test for zero-inflated mixture Poisson regression and give a procedure of component selections based on several criterions. And then the mixture model can be estimated using Expectation Maximization algorithm and be made inference using bootstrapping approach. We demonstrate the advantage of our approaches using the example which motivated this work.
|
