What you would want is a survey where the sample is randomly drawn from a cross section of professionals in the field. This was an online self-reporting survey where anyone was invited to respond. In that case, it is my conjecture that women who have been harassed or assaulted will be more likely to reply, because they will want to get their experience known to the world, whereas someone who has never had an issue may think taking a survey is a waste of time. I'm not a statistician, so I can't really say how to design a survey with a low margin of error, but I know enough to say that this isn't a way to do it. Anyway, I'm glad when issues of abuse are addressed, but when we overestimate the problem, I think it's easy to start to characterize a problem as an epidemic, and then to demonize scientists in the field, when really we don't know what the numbers are.