I know that this is one of those issues around here but this piece on Slate today on the recent new survey is just really good and worth the read when it comes to stats and surveys
From the article:
"This week, I spoke to David Cantor, co-principal investigator of the AAU study. He explained that the 27 campuses he and his colleagues looked at are not nationally representative; the set of schools was relatively large, but it was not randomly selected from the full complement of American universities. While 150,000 students filled out the survey, it was offered to almost 780,000 students, which makes for a disappointingly low response rate of around 19 percent. That, too, is a problem, Cantor said, because it raises questions as to whether those students who did take the survey were more inclined to have been victims of sexual assault, thus inflating the results. He said there is evidence this is the case, explaining that analysis of the survey turned up “some indication that people who did not respond were less likely to be victims.”
This is really not anything new. As a general rule, surveys are extremely self-selecting. This is one of the first lessons you learn in any basic statistics class.
Is there any correlation between the response rate at a particular college and the reported rate of sexual assault at that college? Some colleges tried to entice students to answer the survey by paying them, or by offering a raffle prize. Do those schools have a lower reported rate of sexual assault?
These kinds of surveys are like the comments sections of news articles; the only people who bother to register and comment are the ones who really have something to say. People are so busy, including college students, that they don’t like to bother with something they think hasn’t affected them or won’t affect them. So, yes, the results will be skewed.
OK, let’s suppose that everyone who was raped responded to the survey. It’s reasonable to think that most of them did. Let’s also suppose that the number of people who think it’s fun to lie about being raped on these surveys is small. So instead of 1 in 4-5, it’s 1 in 20-25. Is anyone comfortable with that? There will always be a few psychopathic creeps lurking out there who will attack a few college women, but even 1 in 20-25 goes way, way beyond that. There is plenty of rape culture on campuses. Until it’s no longer acceptable to promote rape culture, and related demeaning behaviors like hazing, nothing will change.
An additional problem that I see with these surveys - and I have looked a little deeper into the surveys to learn what the considered to be a positive response - is they tend to lump too wide a range of contact as a positive response. They will run a headline that states “30% of college women have been victims of sexual assault” and then you find that they include forced sexual intercourse as well as “unwanted touching”, the latter being a little vague and could include something akin to “copping a feel”. I am not condoning feel copping, but I strongly believe it should not be lumped in the same category as rape. But I see this type of survey manipulation all the time. 99% of illegal immigrants are guilty of murder or some other crime!!! (Do you see what I did there? Being an “illegal” immigrant makes you guilty of a crime, by definition)
@mathyone, I see your point and no, in my opinion NO amount of ANY crime is acceptable.
I just don’t think we can assume anything from these types of surveys…can’t assume it was really 1 in 20 or 25, can’t assume it was really 1 in 1000 or 10,000…just as we can’t assume it isn’t really 1 in 2.
I think we can all agree it’s a problem, but the magnitude of the problem isn’t calculable by this type of survey IMO.
OK, referring specifically to rape, according to the article, “for the 2014–2015 academic year, the rate of such completed, physically forced encounters among female undergraduates was 1.3 percent, and for those reporting such acts during the entirety of their college years it was 3.2 percent.”
but this piece on Slate today on the recent new survey is just really good and worth the read when it comes to stats and surveys