To apply mash, the analyst must first conduct a condition-by-condition analysis to obtain an effect estimate and corresponding standard error for each unit in each condition. These estimates are the inputs for a two-step Empirical Bayes (EB) procedure: (1) learn patterns of sparsity, sharing and correlations among effects from the condition-by-condition results; (2) combine these learned patterns with the condition-by-condition results to produce improved effect estimates and corresponding measures of significance. These steps are summarized here, and in Fig. 1. (See Online Methods for more details.)
