Related Experiment Videos
The analysis of categorical data from cross-over trials using a latent variable model.
1Department of Applied Statistics, University of Reading, Whiteknights, U.K.
Statistics in Medicine
|October 1, 1991
Summary
This study introduces a simplified log-linear model for categorical data analysis, particularly effective for cross-over trials. The method efficiently handles nuisance parameters, enabling routine application with standard statistical software.
Area of Science:
- Statistics
- Biostatistics
- Clinical Trial Design
Background:
- Latent variable models are complex for categorical data.
- Nuisance parameters often complicate analyses.
- Cross-over trial structures present unique analytical challenges.
Purpose of the Study:
- To develop a simplified latent variable model for categorical data.
- To eliminate nuisance parameters using conditional likelihood.
- To facilitate the application of log-linear models in cross-over trial analysis.
Main Methods:
- Utilized conditional likelihood to eliminate nuisance parameters.
- Formulated a log-linear model tailored for cross-over trial structures.
- Employed the Generalized Linear Interactive Modelling (GLIM) and SAS statistical packages for analysis.
Main Results:
- Demonstrated a significant simplification in model formulation for cross-over trials.
- Showcased the routine applicability of the proposed method using standard software.
- Successfully applied the model to data from a primary dysmenorrhea cross-over trial.
Conclusions:
- The conditional likelihood approach simplifies latent variable modeling for categorical data.
- Log-linear models are effectively applied to cross-over trials using this method.
- The approach is practical for routine analysis in clinical research settings.