> I can't turn an estimated -5 into positive 500 just because I want to. I didn't have quite that much precision in mind. More like "flip the coefficient's sig

> I can't turn an estimated -5 into positive 500 just because I want to. I didn't have quite that much precision in mind. More like "flip the coefficient's sign" or "make the coefficient bigger" or "make it statistically significant", which are all often doable with enough experimentation (as much as I hate to admit it).