Agreed. The paper treated having a Microsoft-only sample as being good since it's a "large sample size", but the skill set of an engineer can be perceived as go

Agreed. The paper treated having a Microsoft-only sample as being good since it's a "large sample size", but the skill set of an engineer can be perceived as good or bad at different sizes of companies, and not addressing that seems like a huge oversight.