Related Experiment Videos
A program running under MS-DOS for the analysis of epidemiologic stratified or matched data
J C Thalabard1, G Plu-Bureau, B Falissard
1INSERM U. 292, Hôpital de Bicêtre, Le Kremlin-Bicêtre, France.
Computer Methods and Programs in Biomedicine
|March 1, 1989
Summary
Breslow and Day
Area of Science:
- Biostatistics
- Epidemiology
- Computational Statistics
Background:
- Stratified samples are common in epidemiological studies.
- Breslow and Day's method is crucial for logistic regression with stratified samples.
- Current computational demands limit its application on microcomputers.
Purpose of the Study:
- To present an efficient algorithm for Breslow and Day's method.
- To enable the implementation of this method on microcomputers.
- To overcome computational limitations of the standard permutation approach.
Main Methods:
- Developed a novel algorithm using binary representations of permutations.
- Implemented the algorithm in FORTRAN for broad accessibility.
- Focused on optimizing the permutation generation process for logistic models.
Main Results:
- The new algorithm significantly reduces computational time for Breslow and Day's method.
- The FORTRAN implementation is suitable for microcomputer environments.
- Successfully overcomes the limitations of traditional permutation generation.
Conclusions:
- The proposed algorithm makes Breslow and Day's method computationally feasible on microcomputers.
- This enhances the accessibility of advanced logistic regression techniques in stratified analysis.
- Facilitates more widespread use of accurate statistical modeling in research.