Fig. 6.

Performance can be greatly improved by combining data from multiple replicates. Spearman’s rank correlation coefficients and MAE between inferred and true selection coefficients, averaged over 200 simulations with same setup as shown in Fig. 1a, are shown (a,c) when using a single replicate, and (b,d) when combining 20 replicates. The inference using true covariance, and using our estimated covariance (with or without regularization), are dramatically improved when combining multiple replicates’ observations. In contrast, the performance of SL does substantially improve with additional replicates.