I've worked with a decent variety as well, though nothing close to engineering. >Self-reported survey data is notoriously bad This is my least favorite type o

I've worked with a decent variety as well, though nothing close to engineering. >Self-reported survey data is notoriously bad This is my least favorite type of data to work with. It can be incorrect either deliberately or through poor survey design. When I have to work with surveys I insist that they tell me what they want to know, and I design it. Sometimes people come to me when they already have survey results, and sometimes I have to tell them there's nothing reliable that I can do with to. When I'm involved from the beginning, I have final veto. Even then I don't like it. Even a well designed survey with proper phrasing, unbiased likert scales, etc can have issues. Many things don't collapse nicely to a one-dimensional scale. Then there is the selection bias inherent when by definition you only receive responses from people willing to fill out the survey. There are ways to deal with that, but they're far from perfect.