xiechengyong123 / random-metal

Adding random effects model to the METAL software

METAL is a tool for the meta-analysis of genome-wide association studies.

Documentation is available at: http://www.sph.umich.edu/csg/abecasis/Metal/

This version makes two changes:

1. It implements the DerSimonian-Laird random effects model. This requires a third pass through the files, calculating tau-square and the random effects parameters, so it takes about 50% longer than the heterogeneity analysis, and three times longer than the fixed effects analysis
2. The results are now all printed to 7 significant figures (previously it was only 4)

To run the random effects model it is necessary to issue the following command:

ANALYZE RANDOM

This returns

• Fixed effects: Effect, StdErr standard error, pval p-value
• Heterogeneity stats: Direction, HetISq I square statistic, HetChiSq Q statistic, HetDf Q stat DF, HetPVal, p-value for heterogeneity, tausq tau squared
• Additive random effects (DerSimonian-Laird estimator): EffectARE, StdErrARE, PvalueARE
• Multiplicative random effects (The Effect size for this is the same as the Fixed effects): StdErrMRE, PvalueMRE

See the examples/GlucoseExample/metal2.txt for an example.

Note that currently there is only a compiled executable available for Linux machines

If you use this please cite using the DOI above as well as the original METAL software here https://doi.org/10.1093/bioinformatics/btq340