Would be fun to weight each language by average number of stars, but normalize by repository count.

Data analysys without adjusting groups by popularity is a bit lame.

What statistic are you proposing? Number of repos / avg stars ?