Figure 6.—

The relative importance of different summary statistics for the detection of selection under a fixed value of θ. Under different selective scenarios, we investigate the relative importance of our summary statistics. One way of measuring their importance is in terms of the absolute value of the coefficients given to the summary statistics by the boosting classifier. A large coefficient means that a certain statistic is very influential at the considered position for our classifier. Each graph is based on an average of 10 trials, with each trial containing 500 neutral (or bottleneck) samples and 500 selection samples. All the samples were generated with fixed θ. The relative importance of the six summary statistics was considered separately for each subsegment; that is, each time a boosting process was applied to only six statistics at a specific position.