Detecting Outliers in Meta-Analysis
Author Information
Author(s): Freedom N Gumedze, Dan Jackson
Primary Institution: University of Cape Town
Hypothesis
Can a random effects variance shift model effectively identify and accommodate outliers in meta-analysis?
Conclusion
The proposed methodology provides a means to identify and downweight outliers in meta-analysis without eliminating them from the analysis.
Supporting Evidence
- The methodology allows for the identification of outliers without removing them from the analysis.
- It provides a robust approach to handle outliers in meta-analysis.
- The study applied the methods to three contrasting meta-analytic data sets.
Takeaway
This study helps researchers find and handle unusual results in their data so that they don't mess up the overall conclusions.
Methodology
The study used a likelihood ratio test and a parametric bootstrap procedure to identify and downweight outliers in three meta-analytic data sets.
Potential Biases
The methodology does not eliminate outliers, which could still influence results.
Limitations
The methods may not be suitable if there are many apparent outliers or if the collection of trial results is truly unusual.
Statistical Information
Statistical Significance
p<0.05
Digital Object Identifier (DOI)
Want to read the original?
Access the complete publication on the publisher's website