Table 4: Summary of best performing statistics by AUC. Here “model” is the true model used for simulation. Statistics were evaluated by their ability to identify the top 1%, 2% and 10% of genes with large signals (labeled Signal) and large signal-to-noise ratios (S/N). The rows labeled MM correspond to parameter esti-mation using the method of moments while those labeled TT correspond to using the true hyperparameter. Instances where the best performing statistic did not match the appropriate true model and goal are high- lighted in red. Instances where the best performing statistic did not match the model but was consistent with goal are highlighted in blue. A wildcard * indicates the statistic from either the conjugate or indepen- dence model was the best performer. For example, ”*C.F” indicates that the CC.F and IC.F statistics were roughly equivalent best performers.