Table 3.

Scores of homogeneity, completeness and V-measure for the clustering result at different values of the desired number of clusters, k

	Homogeneity	Completeness	V measure
k=2	0.00008	0.00014	0.00010
k=3	0.00142	0.00116	0.00128
k=4	0.13125	0.07194	0.09294
k=5	0.12335	0.06106	0.08168

The value k=4, gives the highest homogeneity, meaning that patients in each group are more likely to share the regimen they were randomised to. Adding a fifth group decreases both homogeneity and completeness because now there are three groups that cannot differentiate well between regimen-C and regimen-DI. In general, additional groups might increase homogeneity but will always decrease completeness.