Related Experiment Videos
On power and sample size calculations for likelihood ratio tests in generalized linear models.
1Department of Management Science, National Chiao Tung University, Hsinchu, Taiwan 30050, Republic of China. gwshieh@cc.nctu.edu.tw
Biometrics
|December 29, 2000
Summary
This study extends generalized linear model power and sample size calculations to handle various covariate configurations. The new method simplifies computations for the likelihood ratio statistic while maintaining accuracy, confirmed by simulations.
Area of Science:
- Statistics
- Biostatistics
- Statistical Modeling
Background:
- Generalized linear models (GLMs) are widely used in statistical analysis.
- Accurate power and sample size calculations are crucial for designing robust studies.
- Existing methods may have limitations in handling diverse covariate configurations.
Purpose of the Study:
- To extend existing power and sample size calculation methods for generalized linear models.
- To develop an approach that accommodates both finite and infinite covariate configurations.
- To simplify the approximation of the noncentrality parameter for the likelihood ratio statistic.
Main Methods:
- Direct extension of the Self, Mauritsen, and Ohara (1992) approach.
- Modification to accommodate varying numbers of covariate configurations.
- Simplification for approximating the noncentral chi-square distribution's noncentrality.
Main Results:
- The proposed method successfully extends power and sample size calculations for GLMs.
- The approach handles both finite and infinite covariate configurations effectively.
- A computationally simpler yet accurate approximation for the likelihood ratio statistic's noncentrality was developed.
Conclusions:
- The enhanced approach provides a more versatile tool for power and sample size calculations in GLMs.
- The simplification reduces computational burden without compromising accuracy.
- Simulation studies validate the method's accuracy across different model and covariate scenarios.