Fig. 3

The average change in improvement from Fig. 2 as a function of the size of the initial clustering, N. For each value of N, each gene family that gives rise to an initial clustering of size less than N and one of size greater than N is considered. The improvement between those two clustering sizes is measured and averaged over all such gene families. For large values of N, the number of families that had a clustering of size less than N and also one of size greater than N but less than our maximum of 200 is very small. Thus, for values of N near 200 the number of samples is small and any statistic for those values is therefore susceptible to error