Related Experiment Videos
Alternative analytical methods for detecting matching effects in treatment outcomes
J P Carbonari1, P W Wirtz, L R Muenz
1Department of Psychology, College of Social Sciences, University of Houston, Texas 77204-5341.
Journal of Studies on Alcohol. Supplement
|December 1, 1994
Summary
Project MATCH evaluated multiple statistical models for large alcohol research trials. No single model proved perfect, but each offered unique strengths for data analysis, guiding future study designs.
Area of Science:
- Biostatistics
- Alcohol Research Methodology
Background:
- Project MATCH provided a dataset for evaluating statistical models.
- Large-scale alcohol research trials generate complex data.
Purpose of the Study:
- To assess the strengths and weaknesses of various statistical models.
- To provide guidelines for statistical model selection in alcohol research.
Main Methods:
- Evaluation of multilevel models.
- Analysis of event history models.
- Assessment of structural equation modeling.
- Review of time series models.
- Examination of ordinal repeated measures designs.
- Application of generalized estimating equations.
Main Results:
- No single statistical model was identified as universally optimal.
- Each evaluated model demonstrated specific advantages for data analysis.
- The study highlighted the complexities of applying statistical models to trial data.
Conclusions:
- Statistical model selection requires careful consideration of study objectives and data characteristics.
- Further research is needed to refine the application of these models.
- Findings aim to assist alcohol researchers in planning and executing studies.