Post 9 (incomplete)

It's On Us. This is a new initiative taken by the President of the United States, the White House, numerous celebrities, many media conglomerates, and major sports organizations, such as NCAA, to tell the people of the United States that they can do more to not rape--on college campuses.

The emphasis is on sexual assault on college campuses. Not in general, not off of the campus, not in prisons, and not in the military, but on college campuses. Why the focus there? Well, because 1 in 5 women who attend a 4-year program will be victims of sexual assault. 1 in 5. That's 20% of all your female friends. Err... wait, not 20%, but 19%. At least, that's what the College Sexual Assault Study (CSAS, 2007; .pdf link) tells us. Well, not really tells once you look at the data.

Now, the emphasis from It's On Us and others who use the CSAS as clear evidence of either "rape culture" or pervasive nature of sexual violence is frequently on the experiences of women. I will point out the flaws and issues with this study and the current cultural view on the matter. I would like to be clear: male victims are sexual assault face equally great challenges as well. They too should be considered in any effort to reduce the incident rates of these violent crimes.

The CSAS is a wretched study, wrought with significant methodological and statistical problems. Allow me to briefly show where:

1. The study looked at only two major universities in the United States. I haven't been able to find which two, but two is a small number. A report by the Institute of Educational Sciences lists the number of 4-year degree schools in 2007 at 2,629. That means the study looked at 0.00076% of all the colleges in the United States. It's not a leap to conclude that this is not a representative sample of the college population.
2. The samples from the universities were derived from the entire enrolled population at the school. Estimates were made on the likely response rates (based on other studies of similar nature and methodological studies), and were then truncated down. That is, even when working with a minuscule sample size of the entire student population in the United States, the researchers then determined the need to cut down the sample pools to a smaller number, based on the belief that too many survey invites would reduce the total response rate. Thus, making the study and even less representative.
3. Response rates were extremely low for this study. After truncation, the CSAS invited 12,836 women to respond. Of that, 5,446 women responded. That's a response rate of 42.4%. This is a considerably low response rate, one which greatly increases your confidence interval (i.e., less confidence in the data).
4. The nature of response rates is challenging, as reasons for nonresponse can vary greatly, depending on the study. Consider, however, the emotional salience of sexual assault: if you were the victim of a car accident, would you be more or less likely to respond to a survey about intersection safety? If you were the victim of a sexual assault, would you be more or less likely to respond to a survey on sexual assault? It stands to reason the transparent goal of this survey (i.e., collect data on sexual assault rates on campuses) would be clear to the participants, and those participants would have the opportunity to respond, if they so choose. Those who see the value in telling their story would probably be more likely to respond to this survey than those who do not. We can reason the likelihood of a skewed response pattern from the low response rate to favour those who were victims of sexual assault or who are close to those who are victims of sexual assault.
5. The numbers used for what constitutes sexual assault in the CSAS are not well organized or explained. The authors report 19% of women who responded (see above for issues with this) were sexually assaulted since entering college. On page 65 of the study .pdf, there is a table that lays out number for the respondents. Importantly, the 19% is at the top, and the numbers are broken down from there: 
• N = 5,446; number of completed or attempted sexual assaults (n) = 1,073.
• Completed sexual assaults = 782 or 13.7% (with no accounting to where the other 292 individuals went in their math, considering the number for attempted was 682).
• Completed is broken down into Physically Forced (256 or 4.7% of the respondents) and Incapacitated (651 or 11.1% of the respondents). So, parent group stated n = 782, but the two child groups are n1(256) + n2(651) = nT(907).
• We find the legal definition of rape (which is defined differently from sexual assault) under the Physically Forced category, which 181 or 3.4% responded reported to have been victims of, and under the Incapacitated category at 507 or 8.5%.
6.  The definitions, as used above, are extremely broad. The study included "forced touching of a sexual nature" (page 52 of the pdf) into the same meta-level analysis as physically forced rape. While the authors do note their analysis does parse out the differences, their results (pages 64, 66, 90) are the only place which clearly state this statistic. The conclusion omits this particular line (a rather curious omission). If the fact that "1 in 5 women are sexually assaulted" was well founded in this (heavily flawed) study, why not put it in the most important part of the paper?!

This post is different from the rest as I will be going back to it to continually edit and add to it. My aim is to thoroughly debunk the CSAS paper in this post and point people to this post as a resource as to why citing "1 in 5 women will be sexually assaulted on college campuses" is both harmful and a blatant lie. @nrokchi