St. Louis Post-Dispatch drops George Will from rotation

daemon is still correct about one thing. You are making up numbers of false positives. You don't have data to back you up there.

 

As many have said, those numbers up there are poor at best, but if you run them through some formulas they seem even more silly.

So lets imagine 1 percent of sexual encounters end in a false-positive report as doctorquant suggest. Well first, since those are reports and 90 percent don't report, we'd have to assume 10 percent of all encounters are "incorrectly interpreted" as assault.

This seems high, but we'll go a step deeper.

Now we'll ponder that someone can regret or feel poorly about an encounter, but not see it as assault. There are obviously no hard numbers. But we'd imagine it would be conservative to say 1-in-4 or 1-in-5 regretted-yet-consensual encounters would be interpreted as assault (I'd say that percentage should be lower, but that factors in later). So we do that math and we're saying that 40 to 50 percent of these women's sexual encounters come to be something they regret.

I assume imaginary U has a notable number of people in happy committed relationships, i.e. a much lower portion of the encounters there could fall into the category of "regrettable" in the short term (not counting a breakup and regretting the years wasted with the wrong person, so just regrettable as an encounter in itself). So that might drive the number of regrettable encounters for single women above 50 percent. So by extrapolating these numbers, we could conclude more than half the sex the average non-committed woman has at this college (and it's only twice a year) comes to be something they truly regret. Now I'm not a women, but the idea that such a high portion of encounters would lead to such bad feelings would make me wonder why these imaginary folks are having sex at all. Maybe there's a puritanical bias in there that women should feel a sense of shame in sex. Or the numbers are out of whack. That could be it.

 

Perhaps Quant should invest more time in learning how to Quallify.

 

Perhaps a terminology clarification might be in order here ... The right-hand column, that's not instances of female students having sex. That's instances of female students having sex with someone new/unfamiliar. I would assume today's college students have sex more frequently than twice a year (would that that had been the case when I was a college student), and I would further assume that in "relationship sex" (for lack of a better term) there's no chance of an assault misunderstanding. That's probably not true, but that brings me to an important point.

What I've provided is a model. Models are abstractions of reality, created to give us an appreciation for the underlying dynamics of complicated phenomena. A given model does not have to be entirely, or even largely, accurate to be informative.

My model shows you how sexual assault cases -- the ones actually reported to and, possibly, adjudicated by these college/university panels -- come from two different sources. Because one source dwarfs the other in sheer size, the number of cases it "generates" can be relatively high even if the rate at which it "generates" these is very low. This is a textbook case of the use of Bayes' Theorem. Here's a simpler example:

Suppose that 1% of the population does drugs. Further suppose that, if administered to a drug user, a given test will come back with a positive result 95% of the time. If administered to a non drug user, this test will come back negative 99% of the time. Now, suppose we have a given positive result in front of us. What is the likelihood that it came from an actual drug user? About 49%.

If I am sitting on that panel, looking at an accusation of sexual assault, it could well be the case that a substantial proportion of the cases I'm looking at are false positives. The downer is that I'll never know whether a particular case is one.

 

Your math/application is so garbled it is impossible to understand your thought process.

Again, according to your figurin', every casual sexual encounter carries a 2 percent risk of a woman correctly reporting sexual assault, and a 1 percent risk of a woman incorrectly reporting sexual assault. If you believe that to be incorrect based on the numbers you provided, please show your work.

Regardless, your numbers are bunk. And keep in mind you are arguing with somebody who agrees with your underlying premise, or at least the spirit of it. As for the report, I have no informed opinion on their methodology except that it is the methodology that resulted in the 1 in 5 statistic that you have based all of your figurin' on.

 

The math is still misapplied since 100 percent of people who "wrongly come to the conclusion" report. The reporting rate is still 10 percent. So either there are 4,000 interpreted assaults, 1 in 10 reported. Or only 40 false positives are reported. Those are still objectively terrible odds 125-to-40, though less terrible than the original interpretation. (I see this was gone over above, oh well)

But lets go back to the original math. Based on the number when there were 400 "false reports" that would mean there were 4,000 false positives. Meaning a woman is 3.2 times more likely to "incorrectly" believe she was assaulted than actually be assaulted (that strikes me as way high). If we assume the 400 represents total "false positives" then only 40 are reported. Which still means a woman is only 2.8 times more likely to be assaulted as to believe they were assaulted by a new partner. Perhaps your point is to tell us someone is driving that rate up. But it also means the model (arbitrary as it is) seems be based on thinking so poorly of women, that in the best light, they can't correctly identify a sexual assault they themselves are suffering at a rate better than 75 percent.

Sorry, I'll need something better than spitball math to make such a bold claim.

(And that's to say nothing of a smaller math flaw in the first half of the model. You took all 5,000 women of the assault pool out of the gen pop. For some reason, the 80 percent of the 1/5th who will be assaulted but not this year are having no sex at all and the assaulted group only has sex with one new partner total. This seems like an odd concession at best. The clarification also makes the assumption someone can't be raped by someone familiar, which is also untrue.)

 

In the original table there are 41,250 sexual encounters. In the original table there are 525 reports of sexual assault. Please help me with this, because obviously your math skills are way stronger than mine ... How is it that when you divide 525 by 41,250 you come up with 0.03? Because that's not what I come up with.

 

What would a George Will victory dance look like?

Re-reading this column about victim advocacy and the rush to judgment ... uh ... yeah.