Related Experiment Videos
Effect-size indices for dichotomized outcomes in meta-analysis.
Julio Sánchez-Meca1, Fulgencio Marín-Martínez, Salvador Chacón-Moscoso
1U Murcia, Dept of Basic Psychology & Methodology, Faculty of Psychology, Murcia, Spain. jsmeca@um.es
Psychological Methods
|December 11, 2003
Summary
This study evaluated effect-size indices for meta-analyses combining continuous and dichotomized outcomes. Two indices, probit and logistic transformations, demonstrated good performance for estimating standardized mean differences from 2x2 tables.
Area of Science:
- Statistics
- Psychometrics
- Biostatistics
Background:
- Meta-analyses often include studies with continuous and dichotomized outcomes.
- A unified effect-size metric is needed to integrate diverse study designs.
- Standardized mean difference is a common effect-size measure.
Purpose of the Study:
- To evaluate the performance of seven effect-size indices for estimating population standardized mean difference from 2x2 tables.
- To compare indices based on bias and sampling variance under normal and nonnormal distributions.
- To identify optimal effect-size indices for meta-analyses with mixed outcome types.
Main Methods:
- Monte Carlo simulation was employed to examine index performance.
- Seven different effect-size indices were assessed.
- Simulations considered both normal and nonnormal data distributions.
Main Results:
- Two effect-size indices exhibited favorable performance regarding bias and sampling variance.
- These effective indices were based on the probit transformation and the logistic distribution.
- Performance was evaluated for estimating standardized mean difference from 2x2 tables.
Conclusions:
- The probit and logistic-based effect-size indices are recommended for meta-analyses integrating continuous and dichotomized outcomes.
- These indices provide a reliable metric for standardized mean difference estimation across different data distributions.
- The findings facilitate more robust and comprehensive meta-analytic research.