Impact of event filtering on random MPRs.

This table shows the accuracy of filtering events from a random MPR, when varying the filtering threshold T. The event supports have been computed from MPRs only, i.e. Δ = 0. Thus, for each line, the set contains all events e from the chosen random MPR having . Column 2 (resp. Column 3 and 4) reports the accuracy as measured by the average symmetric distance (resp. FP and FN) between and . A bold-faced value indicates that the accuracy of the corresponding strategy is significantly better than that of the previous row (p-values of the paired t.tests are lower than ). Column 5 reports the average numbers of predicted, , , and events with or without filtering. On average, the true evolutionary history of a gene family contains 13.7 such events.