We used these simulations to compare mash with several other methods: mash-bmalite, a simplified version of mash that uses only the BMAlite models from Flutre et al5; ash16, a univariate analogue of mash that analyzes each condition separately; and metasoft9,14, which implements several multivariate tests corresponding to different models (the “fixed effects”, or FE9, model, which assumes equal effects in all conditions, the RE2 model9, which allows for normally-distributed variation in effects among conditions, and the BE model, which extends the RE2 model to allow for effects are exactly zero14).
