Meta-analysis inside and outside particle physics: two traditions that should converge?

Area of Science:

• Statistical methodology
• High-energy physics
• Epidemiology

Background:

• Meta-analysis is widely used in medicine and epidemiology since the 1970s.
• High-energy physics has employed meta-analysis for particle properties since 1957 via the Particle Data Group.
• There is limited cross-disciplinary interaction between these fields.

Purpose of the Study:

• To investigate differences in meta-analysis practices between high-energy physics and other scientific fields.
• To evaluate the utility of systematic errors in particle physics meta-analysis.
• To compare heterogeneity treatment methods: error scaling versus random effects models.

Main Methods:

• Application of statistical models to compare meta-analysis practices.
• Development and utilization of an exact test for the error-scaling hypothesis.
• Discussion of methodological divergences between particle physics and general meta-analysis.

Main Results:

• Systematic errors are currently found to be of limited utility in particle physics meta-analysis.
• The conventional random effects model demonstrates utility within particle physics.
• The error-scaling hypothesis was tested and discussed.

Conclusions:

• The standard random effects model is recommended for particle physics meta-analysis.
• Systematic errors do not appear to offer significant benefits in current particle physics meta-analysis.
• Encourages greater exploration of statistical applications in diverse scientific fields.