I recently made a blog post on which universities have the most successful startup founders: http://minimaxir.com/2013/07/alma-mater-data/

In the post, I made it clear where the data came from, how I processed it, and additionally provided a raw copy of the data so that others could check against it.

It turns out the latter decision was especially a good idea because I received numerous comments that my data analysis was flawed, and I agree with them. Since then, I've taken additional steps to improve data accuracy, and I always provide the raw data, license-permitting.