Figure 2:

(a) Boxplots of $R_{\text{pERP}}^2$ from regressing the true pERPs on the estimated pERPs in 100 simulation runs show that the pERP-RED algorithm is correctly identifying 5 as the true number of pERPs. (b) Boxplots of the $R_{\text{test}}^2$ obtained by regressing ERP records from the test set on the estimated pERPs, for both the high and low noise simulation cases. In both cases, the boxplots of $R_{\text{test}}^2$ identify 5 as the true number of pERPs as the value of $R_{\text{test}}^2$ levels off sharply at that value.