Figure 6. Top features most often significant to differentiate files by quality in each of the three subset groups.

For each of the three subset groups (A, B, and C), the number of subsets (y-axis) in which a quality feature is significant is shown. Subsets have a minimum of 10 files and the P-value cutoff for significance is defined as either 0.01 (top row), 0.001 (middle row) or 0.0001 (bottom row). The total number of subsets in each subset group is 436 for subset group A, 354 for B and 461 for C. The number of subsets with at least 10 files in each subset group is 38 for subset group A, 41 for B and 33 for C.