It's week 2 of the pick 'em pool, and now that the picks are published, the question comes up who's picking with the crowd and who's going out on their own. It being impossible to decipher 19 sets of 16 game selections, I pulled out my handy-dandy statistical modeling tool, S-Plus, and whipped up a quick cluster analysis.
First, of the 19 people involved, the data conveniently breaks into 2 clusters, with 14 people picking essentially the same, and 5 going out on their own. Of those 5, 3 are more similar than 2, who are each out there.
The 5 in the smaller cluster are DK, JF, and TM., with TR and (yours truly) DH the 2 off on their own.
The silhouette analysis shows these groupings
The larger top group is shown first, with the length of the bar the similarity to others within the group. Next is the first outlier, then the second group, then the other outlier.
WIth an idea of who's in the in and who's in the out crowd, I turn to the question of who is thinking most like whom. The clustering tree helps with this, putting neighbors side by side. The height is a measure of dissimilarity, so all of the folks in the .1 area are, for the most part, sticking together. 
The tree format is also a measure of similarity, so folks that hang off the or near the same branch have selected alike.
Most similar are TJ and JoB, and looking at their picks for the week, they agree on everything but the Sunday night toss-up between San Diego and New England. SC is thinking much the same as TJ, differing only on the Dallas/Miami game.
All the picks in convenient Excel are here if you want to play around and compare.
Comments