Our method, mash, estimates effects of many “units” (J) in many conditions (R), allowing that effects may be sparse (i.e., many zero effects), and allowing for correlations among non-zero effects in different conditions. For example, in multi-tissue eQTL studies, the “units” are eQTLs (J > 10,000), the conditions are different tissues (R = 44), and mash estimates the effect of each eQTL in each tissue, allowing for cross-tissue sharing and tissue-specificity of eQTLs.
