>Projects from charter, magnet, and KIPP schools, and projects from Teach for America teachers tend to be more successful regardless of gender, and happen to have more male teachers, so <i>we filtered them out of the data to eliminate that confounding effect.</i><p>>We then controlled, one-by-one, for the other variables that were correlated with both gender and project success. For example, men are much more likely to create projects asking for new technology. But technology projects actually tend to be less successful than other projects, so we ruled that out as a driver of the gender difference.<p>I'm not a statistician, and don't have access to all of the raw data anyways, but: You can't filter out data which moves the mean in the direction you want and not filter out data that moves the mean in the other direction. If technology projects lower male teachers' success rates then leaving them in is just as bad as leaving charter schools in.