This is another example of how you actually don’t understand when it is appropriate to use statistics and when it’s not. You don’t just blindly pick characteritics that happen to apply a given situation-- you need to have some idea that the variable has some reason to be correlated to the incident. Is there something in the nature of being a lacrosse player that might make a persson more likely to rape or not? Highly unlikely. One might argue that “being a member of a college sports team” has some relavence, but lacrosse vs soccer vs. volleyball? Unlikely.
You keep talking about how important it is for these types of statistical analyses to be done by law enforcement officials, but it’s unlcear to me if any agencies actually use the type of statistical analysis you’re talking about, except in particularly heinous crimes where there is no information at all about the suspect. I’ve never heard of a profiler being brought in for a single instance of rape when the accuser lists the actual names of the accused. But I’m open to being educated on that if I’m wrong.
And, IIRC, you agreed in an earlier post that what you were talking about was more along the lines of a “gut feel” to be used by cops, rather than some sophisticated data base. And once again, I think the cops need to be careful about using “gut feel”, as our guts are often wrong. Put race into the mix, and I can’t think of any good that will come out of that. Imagine a police force whose comissioner talks about the tehcniques his police use and says something like 'I encourage my officers to use their gut feel to help determine the credibility of certain vicitims. You know, things like using the race of the alleged vicitm and the race of the alleged criminal to figure out how likely it is that the crime actually happened." The mind boggles.
You know what, Huerta? Plain language can go a long way to preventing misunderstandings. If you want us not to misconstrue what you are saying, you should reach for words with the most meaning in them, not those that make you look smart. I like the words you used above and fully understand them, but they don’t make what you are saying more sound or reasonable.
I strongly suspect it’s your way with words that have convinced people you know what you’re talking about.
Oh, gosh, he caught me in a typo or ten, I must, therefore be disbelieved. Sure. Couldn’t possibly be anything like arthritis in my hands, visual problems, simple typos and/or a historically bad spelling problem, none of which would be germane to this discussion (of course that hasn’t stopped you before).
but even if the data point is “once a week in the state” vs. the city, it still ain’t ‘vanishingly rare’ to the point that some individual accusation should be questioned out of hand because of it.
and I doubt that any of your prior supporters would agree w/you if they realized that you discounted that a particular instance happened even though it happens once a week in that state.
wring and that’s real question isn’t it? There was plenty of doubt in this case, without digging up stats on interracial rape. Yet it was done. Cherry picked and used as a club.
Yes, yes, we were told to ignore, to forget about them; not because the stats didn’t apply and it was a mistake to use them; but because we couldn’t handle the truth…that whites, just don’t rape blacks. “Sorry”, he said…with glee.
I submit this, had Jeese Jackson dug up race based stats to disprove the reliability of the DUKE players, he would be called a race baiter, a racial racketeer and an opportunist. Had he stated that his detractor’s were upset that there wasn’t more raping blacks, by whites; he would have been shown no mercy in this thread.
I fail to see why Huerta88 should be treated any differently…barring any other evidence; of course.
not when you use different methodology in your preferred stats. As I’ve pointed out, you picked out the one broader category for w-o-b rape (national vs. any other smaller subset, general rape vs. gang) when that data supported your contention, but when the national vs. smaller subset did not support your contention, you insisted that the smaller subset be more significant. IOW - your choices seemed dependant **not ** on some distinction that may have relevance, but rather on 'did the data support my (preconceived) belief
you ‘rejected’ it by focusing on teh smaller subset (with the very very very small single study) whenever discussing the concept. “up to 50%” was your mantra, even though the much larger dataset showed remarkably smaller number. That’s not even touching the concept of relying on a single small study to make any kind of generalizations in the first fucking place.
Wasn’t it you who brought up the MacMartin case? A study of that particular case done at the time of the trial would have shown a remarkably high number of child victims of Satanic ritual, yet we (hopefully) would reject conclusions and projections based on that single case study. I don’t know the individual who did the case study you constant refer to, and am not accusing them of intentional error. but their results do significantly differ from national data, without really addressing why that might be.
Nice try. If you had walked into the thread with a different bias, you might have come up with something else.
We model the occurrence of white-on-white rape with one Poisson process and white-on-black with another. The first has a higher rate (for those unacquainted with the math, whenever I say “rate” I mean average number of occurrences per woman per unit of time) than the second. We model false accusation with a third Poisson process and assume, in the absence of data, that whites and blacks are make false accusations at identical rates. Result? A priori, a black woman claiming to be raped by a white man in this model is less credible than a white woman claiming the same.
I have no idea which of our models is closer to the truth. Since no one’s found a racial breakdown of false rape reports, it might be interesting to compare the rate at which various crimes occur to the rate at which they are falsely claimed to have occurred.
Why a poisson? Why not a negative binomial? More to the point, why do you feel that describing the shape of the distribution with a Poisson model will allow you to make any conclusion about credibility? You’ve not advanced the ball at all, since all your models will allow you to do is say that in this set, one type of rape is reported more frequently than another. It has no bearing on the credibility of the report of a separate individual.
Besides, aren’t you really saying that if you have made a priori assumptions and have no data, all you’ve got is your assumptions? Why all the jargon? What is your point?
You’ll note that there was no model assumed in my hypothetical. I was dealing with raw numbders and no assumptions (except the statistics for reports of past rapes). And there was specifically no modeling at all of false accusations of rapes because we were assuming that we knew nothing other than the race of the individuals. The whole idea is that we have no information about that and no way to make any inferences. You obviously having been following that point, and so I’m afraid you don’t know what you’re talking about.
If Huerta88 ever digs up data to the effect that reports of white-on-black rape are significantly more likely (in the statistical sense) to be false than those of white-on-white, I see no reason why he, as a private citizen who is hypothetically not aware of any other facts about the situation, shouldn’t put more faith in the latter.
Just to be clear, this goes for any comparisons of X-on-Y rape to Z-on-W, for any values of X, Y, Z, and W. It’s unfortunate that Huerta88 has centered this discussion around race without evidence that race is a significant factor in the relative frequency of false reports, but I for one am not willing to make an exception. If that makes me a racist, I will gladly be a racist to the end of my days. If I were the district attorney or a juror on this case, I would do my utmost as a logical human being to suspend my judgment while the facts are presented.
What part of “not knowing anything other than the race of the people involved” do you not understand? That’s the issue. Assumptions were being made not only without the statistics on false reports, but with a statement that we didn’t need to know anything about them to make a judgement against the woman in question.
My point is that it’s really, really easy to reach any conclusion you like when you start making what seem like equitable assumptions. I will say more when I respond to the post right after yours.
Why Poisson processes? They’re frequently used to model this sort of thing, and there seems to be enough statistical expertise for the reference to be appreciated. A random choice, really, but it saved me the trouble of spelling out the requisite independence assumptions.
I’m sorry, but that’s not what you said up-thread. Let me repeat:
Emphasis mine. His justification is incomplete, but that doesn’t make his theory wrong. You gave a list of scenarios, some of which made unjustified assumptions themselves, in an effort to show that, under equitable hypotheses, Huerta’s theory had to be incorrect. In order to even things out, I proposed a scenario in which he could have been right, under similarly equitable hypotheses.
It’s unfortunate that while mathematics always has the look of ultimate truth about it, mathematical modeling (which I take to include statistics) is incredibly subtle.
Really? I don’t get the sense that many people here would really appreciate discussions of various families of distributions. I kinda thought you might be trying to throw around a bunch of jargon to support your conclusion in the hopes that it might sound impressive. Of course, I tend to assume such base motives too often. Perhaps that is not fair.
But, gee, let’s go through the possibilities for general linear models, since so many people would love to dig their teeth into it. I would expect that you would see significant overdispersion in the distribution of “white on black” rape, so I would think a negative binomial model would be more appropriate than a Poisson model.
None of this, of course, has a whit to do with credibility. It seems to be some smoke and mirrors, but there I go again.
If you believe that one is justified in lending more or less credence to a rape claim based on statistical properties of an ensemble of claims like it, then I have no beef with you.
One group of posters has been making arguments that ignore the conditional probability of a given claim being false. These are wrong for reasons obvious to anyone who has mastered elementary probability. Another group has been arguing that statistics should have no bearing on anyone’s belief in the veracity of an “individual” claim. That’s like saying “I shouldn’t be sure that the sun will rise tomorrow, because tomorrow is a new day.” I’m pretty sure no one can come up with a workable epistemological theory without some mechanism like statistical inference.
Because of the racial overtones throughout the thread, I want to make perfectly clear that while I hold beliefs that are “racist”, such as “my pension is much more likely to be stolen by a white man than a black man”, I try to be exceedingly mindful of how they are reflected in my actions.
[QUOTE=bored mathematician]
If Huerta88 ever digs up data to the effect that reports of white-on-black rape are significantly more likely (in the statistical sense) to be false than those of white-on-white, I see no reason why he, as a private citizen who is hypothetically not aware of any other facts about the situation, shouldn’t put more faith in the latter.
Okay, now I’m fairly confident that you are throwing around terms you just aren’t too familiar with to try to influence others regarding your argument. I say that because it’s fairly clear that you have no idea what you are talking about.
A lifetime of reading and writing mathematical prose has left me precise to a fault, often quite needlessly.
My point was very simple. John Mace, in debunking an argument made by Huerta, seemed to claim that Huerta’s conclusion was wrong. He proceeded to present three scenarios in which this was in fact the case. I added a fourth scenario to show that while Huerta’s argument not valid, his conclusion might still hold in the absence of further on-their-face inequitable assumptions. It was not intended as a serious model, despite the way I described it.