Thanks Rosebud! This is a cool analysis, and I really appreciate you sharing the code. There are a ton of ways to pick apart the interpretations, but I feel that you were pretty up front about the difficulties of inferring much. I can empathize with the level of data cleaning fatigue you must have felt. Here's an analysis of VC backed enterprise startups that I did for a stats class: http://robterrin.com/MissingData/index#1

By the end, I was pretty much burnt out on this data set to draw many strong conclusions. Still, you busted some myths and flagged some interesting observations for further research. Thank you!

P.S. Oops, I realize the OP is not the author, but cool nonetheless.