Hello,
I would like to perform a meta-analysis on beta slopes from a regression model I use to analyse data from several studies.
However, I occasionnaly encouter some obvious outliers (for instance, a value of 4 on studentized deleted residuals) in a given data set.
Although I am aware of how to manage outliers in a single study, it seems to me that managing outliers in a meta-analysis might follow different rules and "policies".
Could anyone suggest a proper way (or provide a reference) to handle outliers in a small-scale meta-analysis ?
Thanks !
Wojciech
The book by Stanley and Doucouliagos is the best reference I know. Check page 101 where they describe:
"One of the beauties of WLS-FAT-PET-MRA [meta-regression] is that estimates with large standard errors are given little weight, while precise estimates are given a much larger weight. All of the studies with implausibly large elasticities [or beta's] also have correspondingly large standard errors; thus, the WLS-MRA automatically discounts them. MRA's are remarkably resilient to such outliers." They also report on a robust meta-regression model that is even more resilient to such outliers (see their 2099 paper)
Article Publication Selection Bias in Minimum-Wage Research? A Meta-...
Article Meta-Regression Analysis in Economics and Business
Robert and Levente, thanks a lot both of you for your insights !
- Badges
- Science method
- Similar topics
- Mathematics
- Statistics
- Statistical Techniques
- Meta-Analysis