The probable explanation for the absence of statistical significant differences in many of the cells in Figure 7 is the relatively small number of regions in some combinations of D-P and Structural types (see Table 3 below). The IR type is particularly small, with only 23 regions. Combining the IR type with another D-P type might reduce the problem of small sample sizes. On the basis of the pattern of results in Figure 4 it could be argued that the best way to combine types would be to retain the two accessible types (IA and PRA) and to merge.
