Related Experiment Videos
A score test for overdispersion in zero-inflated poisson mixed regression model
Liming Xiang1, Andy H Lee, Kelvin K W Yau
1Department of Management Sciences, City University of Hong Kong, Kowloon.
Statistics in Medicine
|June 24, 2006
Summary
This study introduces a score test for zero-inflated Poisson (ZIP) mixed models, addressing potential bias from overdispersion. The proposed test effectively distinguishes between ZIP and zero-inflated negative binomial models in medical count data analysis.
Area of Science:
- Biostatistics
- Medical Statistics
- Epidemiology
Background:
- Count data with excess zeros are prevalent in medical research.
- Zero-inflated Poisson (ZIP) regression models are commonly used for such data.
- Hierarchical or correlated count data may require ZIP mixed regression models.
Purpose of the Study:
- To propose a score test for evaluating zero-inflated Poisson (ZIP) mixed regression models.
- To assess the performance of this test against the zero-inflated negative binomial alternative.
- To address potential bias in ZIP parameter estimates due to overdispersion.
Main Methods:
- Development of a score test for model comparison.
- Simulation studies to evaluate the test statistic's sampling distribution and power.
- Application of the test to pancreas disorder length of stay data.
Main Results:
- The proposed score test demonstrates satisfactory performance across various conditions.
- Simulation results indicate the test's reliability in distinguishing between ZIP and zero-inflated negative binomial models.
- The test was successfully applied to real-world medical data.
Conclusions:
- The score test provides a robust method for selecting appropriate models for overdispersed, zero-inflated count data.
- This approach helps mitigate bias in parameter estimation for medical count data analysis.
- The methodology is applicable to complex healthcare datasets, such as length of stay in medical disorders.