Well, it’s just that agreeing with this proposition seems to lend some weight to statistical prediction, that’s all. 
But note, Malacandra, that were aren’t talking about stats drawn from a population, but from an individual. A credibility score of 10% for an individual would be enough to justify skepticism, IMHO. However, knowing that one out of 10 guys named Random have credibility scores of 10% would NOT justify skepticism, IMHO.
I’m glad I could turn my little insult into a teaching moment.
As it’s only a humble opinion it’s not much of a teaching moment, really. If experience shows that hornets, as a species, tend to sting people who piss them off, I’m cool with the notion that if I piss this particular hornet off it will probably sting me. Better, delete “hornet” and insert “bee” - given that no individual bee can have a track record for stinging.
Personally I think that some of the people calling other people “stupid” in this and the other thread should consider whether their adversary does in fact have a verifiable track record for stupidity - and if not, should reconsider the label, following the principle that “everyone who disagrees with you isn’t necessarily an idiot”. ([Dogbert]Bah![/Dogbert])
Stinging is inherent to wasps. Stinging is to wasps as walking erect is to humans. So no, it would not be erroneous to judge one wasp on the basis of what other wasps have demonstrated in the past. “Stinging when angry” is part of their makeup, as determined through years of observation.
In terms of conscious, free-willed behavior, there is nothing that people named Random have to have in common. So looking at a population of “Random” people and making conclusions about one individual with this name is wrong in all kinds of laughable ways.
Now replace “Random” with “black women” and “name” with “race”.
Do you see yet the insanity that provoked me to tears earlier in this thread? I’m not going to call you stupid if you don’t, but to me it’s so obvious that I (or monstro or Biggirl or Urban Chic, etc.) should be judged individually and without regard to other people who happen to black, or white, or female. This strikes me as being completely indisputable, but it’s not apparently. And I don’t yet see why it shouldn’t be.
It’s really hard to do this when someone is saying really stupid things that, when put into action, have God awful consequences. And when you try to reason with them patiently, they are fractious and beligerent. That’s sign of stupidity and there is no crime in diagnosing a disease based on its symptoms.
My thoughts:
- “88,” when attached to a username on an Internet messageboard, immediately makes me think “Neonazi.” No other number has similar associations for me when attached to a username on an Internet messageboard, but that’s what 88 makes me think of. It’s based on my experiences on a civil liberties board in the late nineties, experiences that left a strong impression.
1a) I do not, of course, dismiss someone’s posts based on that number; I know, of course, that there are other reasons to use that number.
1b) I accept Huerta’s explanation, and even before I read it, I didn’t think of him as a Neonazi–it just made me do a double-take on several occasions where I read his username. -
Huerta’s challenge is legitimate: even if he’s the most vile person on earth, that doesn’t negate the strength of his arguments (arguments which I find cringeworthily weak). In most cases, it’s preferable to focus on the argument, not the arguer.
2a) However, Neonazis on the Internet are a specific exception to that rule. My experience with them has been that they argue dishonestly: they deliberately put forward things that they know have been refuted, they deliberately twist the words of their opponents, they deliberately lie as part of an effort to show their own cleverness. When someone is arguing dishonestly, it’s a tremendous waste of time to continue arguing with them; it is legitimate to point out dishonesty to others. This is one of the few cases in which an ad hominem attack is relevant.
2b) If we are to imagine Huerta as a neoNazi, then, we must imagine him as a sincere neoNazi (a rare bird indeed), if we’re going to address his posts seriously. Alternatively, if we conclude that he is insincere, his posts are not worth engaging with. I’m not sure whether he’s honest or not, but since I"m not sure he’s dishonest, I’ll treat him as if he’s honest. - Questions about which of two stories (the bet-on-red or the bet-on-33) is more credible are irrelevant, for one key reason: we’re not being asked to choose between two stories. Hell, we’re not even being asked to evaluate the shade-of-grey credibility of one story: at most, all we can legitimately do is to declare one story either beyond-the-pale unbelievable or not.
-
Huerta put forward the DOJ stats (with a flawed interpretation–one that I would also have had, to be fair) to suggest that the rape story in question is beyond-the-pale unbelievable. In actuality, it suggests precisely the opposite: even if you accept his flawed interpretation of the stats, the conclusion is that up to 9 white-on-black rapes have been happening annually in recent years in the US. As such, a report of a specific white-on-black rape is not beyond-the-pale unbelievable: we have a very good reason to believe that they do happen. Bricker’s ninja scenario, therefore, is irrelevant: we have no reason to believe that ninja gangs have ever attacked a liquor store, much less that up to nine ninja-gang-liquor-store-robberies happen annually, much less that helicopters are involved.
4a) I don’t see a fair reason to call Bricker dishonest here–but then, I don’t see a fair reason to call ywtf stupid, either. I think ywtf has put forward the much stronger statistical argument, but I think she’s also accused Bricker inaccurately of believing that his incorrect statistical analysis would be relevant in a jury trial. - I think this would be a great topic for a GD thread, but then, that may only be because I’m coming into it so late. Would folks be interested in moving it there, where the vitriol might be diminished?
Daniel
If you apply this analogy to the prediction you’re making in this thread, it would work thusly:
"If experience shows that (Black women) as a species tend to (lie about being raped by a white man), then it would indeed follow that this particular (black woman) would (probably lie about being raped by a white man).
you keep on forgetting that the factor isn’t really ‘w-o-b rape’ but rather can this particular woman be believed
and my answer is: looks like we shouldn’t believe her in this case 'cause of all the evidence such as time lines, DNA lack thereof, tainted identifications etc, but not because of some years national stat about rape/
I don’t think the main accusation against Bricker is that he is dishonest, unless you define dishonesty as not being able to admit that when you’re wrong and generally being an ass about it.
With respect to Bricker’s analysis, I never said he believed his analysis would be relevant in court. My position was that it’s naive to say that primae facie determinations of credibility based on a faulty thought process wouldn’t translate into bias that unfairly favors one side over another. Bricker stated that the stats would be inadmissible in court, but that doesn’t really impress me much. In the hypothetical posed by John Mace, Bricker said that if he thought that w-o-b rape was “vanishingly rare” (according to nationwide crime stats) that would cause him to view an accusation of such as less than fully credible. Which means that if he was a juror in the Duke case, the prosecution would have to overcome that bias plus a “reasonable doubt” in order to prove its case.
[quote]
LHoD: I think this would be a great topic for a GD thread, but then, that may only be because I’m coming into it so late. Would folks be interested in moving it there, where the vitriol might be diminished?
I would, indeed. But do you think that it or the LW thread would survive without the ability to hurl insulting remarks at black women and ywtf and monstro in particular?
Ellis Dee, your questions was answered better than I was able to put it into words. But folllow along with this weird story for a moment:
-
Woman #1 was married and her husband died. She remarried and she and her new husband had a baby boy named David.
-
Little David grew up and got married. But his first wife died and he then married Anna.
-
Meanwhile, ten miles away, Man #1 had married Elizabeth. But she died.
-
Nearby, Mary had married George, but he died.
-
Both Mary and James had bunches of children but they decided to marry each other and they did. And they had two more children together for a total of 17.
-
The youngest of the 17 was born when his father was in his sixties and his mother was in her mid-forties. Wow! Was he a surprise! His name was lets say “Tom.”
-
Meanwhile, back to Dave and Anna. They had nine children. The 8th child was born when Anna was in her forties too. Her name is “Sassy.”
-
Tom and Sassy met and fell in love and got married. Even though Tom had been told that he couldn’t sire children, he fathered his second child when he was thirty-seven.
-
Given all of the deaths that would have been necessary and the coincidences of time and place, what are the chances that the offspring of Woman #1 and Man #1 would be posting to you now?
I was unable to follow the story, nor have I seen a compelling answer to my question. But I don’t much care as I am well past weary of this irrelevant issue.
astro created it seven weeks ago in an effort to move this unending hijack into its own thread:
Should statistics be used to evaluate the veracity of an alleged victims claims?
Sorry, I thought I was in the Lying Whore thread. astro created the GD thread to try and end the hijack to that thread. Didn’t seem to take. Perhaps people just really wanted to scream at Huerta for another 20 pages, and it’s no fun to try and do that in GD?
And people keep getting emotional and illogical about statistics. As many have said in both this thread and the other:
if there is good evidence to rely on that either supports or disproves the allegation in question, it is much better to go with that than rely on generalities.
But that does not magically make the statistics false or useless.
And another statistic we should be bearing in mind is whether black women, as a class, are more disposed to file false rape allegations than true. My natural expectation would be that if actual rape of black women by white men is rare, complaints of such would be correspondingly rare - in which case, any such complaint should be given no more and no less weight than if it were by a white woman. But I’d rather have the facts than my natural expectation any day of the week; and reserve the right not to be called a racist or an idiot for merely asking for the data and giving it due consideration.
Well, quite. It’s a bit like patiently demonstrating through mathematics that a witness who is really good at telling maroon cars from red ones can still, owing to demographics, have only a 28% chance of being right in her identification, and having this proven fact agreed on, and still having your opponent insist that the witness should be treated as really reliable because, after all, she can tell maroon from red 95% of the time. So diagnose not, lest ye be diagnosed. :dubious:
Unless there are zero cases of black women correctly reporting rape by a white man, then no, this is not a statistic to keep in mind when we’re evaluating such a claim. Keeping it in mind serves no useful purpose, since we know that such claims can be correct; we should evaluate the claim based on the information of the specific claim, not based on generalities that may be irrelevant.
Daniel
On any grounds other than ideological ones? Hypothetically, were there on average fifty false claims for every true one (which I do not in fact believe to be the case), do you really not think there would be more grounds for scepticism than if there were fifty true claims for every false one?
Generalities shouldn’t factor into it at all, because its A) faulty logic and B)presents bias. If you don’t have evidence, why can’t you just wait until you get some evidence before you make any positive or negative assessments? Suspending judgement and all that.
You really haven’t presented any compelling reason why the particular stats in the LW thread bring anything of value to the discussion. Until you do that, then this refrain is uncompelling.
You’re not the only one who has presented this line of argument because even those sympathetic with the “stats are irrelevant” position have put forth a similar idea, and I don’t understand how this is logical at all. Unless all major confounders are tightly controlled for (socioeconomic class, occupation, age, etc) and unless we are only looking at one specific type of rape (such as gang rape, or date rape, or jumped-in-an-alley-by-a-stranger rape), I don’t understand how you can look at all w-o-b rapes and make any determinations about race and false accusations taken down to the individual level.
Because then you’d be making the same type error that I pointed out here, in which you’d be taking statistics derived from a wide array of rape allegations, filed by different types of people in different types of circumstances, and attempting to draw conclusions of likelihood for one individual report based on only one qualification: “college co-ed”. As if college co-eds are a species unto themselves and not simply a group of people who happen to be going to school at higher institution of learning.
Can you (or anyone else) explain the validity of this argument for me?
And you are so sure you right, aren’t you? Even after the Zoe’s random number generator should have told you how illogical your position is.
You argument boils down to this, Malacandra: Anyone who spots an “unusual” car in a hit-and-run is probably mistaken. An “unusual” car is any car that has a 2% prevalence or less in the car population. So I just saw a yellow Mustang bump into a white Honda and drive off. Oh, but I’m probably mistaken because afterall, how many yellow Mustangs are there in this town. Probably a lot less than 2%. That’s such an unlikely event and therefore implausible.
Does this make any goddamn sense? Of course not! It doesn’t matter if the car is maroon, red, or fruscia. That doesn’t change the plausibility of the scenario. When a car in involved in a hit-and-run, the prosecution doesn’t go to the DMV to see how many cars are registered in that town matching that description. Because it doesn’t matter. If a witness says they saw a maroon car collide into another car, and a suspect with a car of that color has been identified, it doesn’t matter how common those types of cars are. If the witness is 95% proficient in spotting the defendant’s car, then they bring evidence to the table.
I don’t see why this is so difficult to understand. You are making the same error that the creationists make. Are you a creationist? This is a (half-way) sincere question.
And around and around and around we go…
Who am I to argue with you on the subject of uncompelling refrains?
Oh, I’m all for getting the most specific and applicable stats we can, believe me. It’s just that you appear to me to be arguing that unless we have every stat, and have also refuted every possibleargument as to why each of them could be wrong, we should pay no attention to them whatever.
You’re not my wife posting under a pseudonym, are you, to spout “you’re so sure you’re right!” as though it were some crushing rejoinder. Can’t recall Zoe’s random number generator so I’ll have to just run with the counter-arguments you’re providing for yourself.
Something of an oversimplification. No-one would seriously argue that you couldn’t tell a yellow Mustang from a steamroller, but they probably aren’t the only choices. Question is, how many yellow cars shaped roughly like a Mustang are there in town, and how likely is the witness to be able to tell a yellow Mustang from the other kind of yellow car?
Just because you italicise the phrase “Because it doesn’t matter” doesn’t mean it doesn’t. The whole point of the mathematical exercise was to demonstrate that, as a matter of hard fact, the witness is not 95% proficient in spotting the defendant’s car. That the witness might have been 95% proficient in some imaginary universe in which all makes, models and colours of car were evenly distributed – which is what the hypothetical lab test showed – does not mean that the witness is actually 95% proficient in spotting a maroon car in the real universe; the far greater prevalence of red cars mean that most supposed “maroon” car sightings are actually red cars misidentified as such.
Now if you have an actual suspect, good for you, but if all you’ve got linking him to the hit-and-run is an identification of his car colour which is rare, but occasionally mistaken for a much commoner colour, such that the unlikeliness of misidentification is outweighed by the rarity of the colour - which does come down to mathematics - then the statistics become extremely relevant indeed.
Doubtless the prosecutor will indeed not stress this point overmuch, but if counsel for the defence has his wits about him, access to the facts, and the good luck to find a jury that understands some not-too-difficult mathematics (perhaps with the aid of a diagram), well, he’s on a roll.
Because I’m right and you’re wrong. Would that I had the New Scientist article to hand, 'cos I didn’t make up this mathematical marvel out of my own head, yanno.
Am I? Well, bless me, that must make me a…
I decline to answer the question on the grounds that you are perpetrating so many logical fallacies in that last paragraph that you’re on the verge of making my head asplode. We can drag in Ad Hominem for a start.
So you think you could get data that targets and/or controls for the following factors*?
- ages of victims and ages of accused
- relationship of accuser to the accused
- presence of drugs
- certain community factors (such as settings that support sexual violence )
- occupation of the accuser (or presence or absence of high-risk sexual behavior)
- type of rape (gang, acquantance, etc)
- socioeconomic background of accuser and the accused
- presence (or absence) of any prior sexual history of abuse
- …
Because unless you can find data that looks only at the specific factors at play in the Duke case, any conclusions you make about the individual taken from data from a diverse population are pretty much useless. It sounds crazy but there’s a lot of variability that exists in the “black woman” cohort. I don’t know why we should treat them as some fungible class of people just because its convenient.
That doesn’t matter. The conclusion you reach about the reliability of the witness ultimately boils down to the “rarity” of maroons, not so much due to their resemblance to reds. But what does it matter if maroons are “rare”? The witness is not being asked to testify about a population of all the cars in the town. They are only being asked about the car they saw that day, at that time, in that area. Not a car randomly selected from the whole population.
Based on your reasoning, there’s no real reason why you should have limited the hypothetical to maroon-car prevlance for just the city. Why not look at a nationwide maroon-car prevalence? Afterall, state lines can be crossed, right? Wow, when you do that the PPV might even get smaller!!! OH MY GOD! ITS UNBELEIVABLE!
Except that its not. This logic is so ridiculous that it’s begs to be laughed at.
They are 95% proficient in spotting a car matching the defendant’s color. That’s what sensitivity is, Malacandra. sigh The witness’ reliability is established by how well they are able to distinguish maroons from reds; the positive predictive value only tells the probability that they will be right when “testing” at random in a population of reds and maroons. But as I said before, the witness didn’t “test” at random in a population. If the defendant was known to be near the scene of the crime, they were driving a maroon car, and a witness said they spotted their car and “look, I’m able to correctly identify these types of cars almost 100% of the time”, then it doesn’t matter how many maroon cars there are in the total population.
But your hypothetical read thusly:
(bolding mine)
So if we have a “key witness” and a “case” that sort of suggests we have more than just some bloke on the street saying things. Or else we wouldn’t have enough to go to trial and the question of “how reliable is this witness” pretty damn moot.
Au contraire - the whole point is that not only are maroons rare but that they and red are occasionally mistaken for each other.
And how does the car that they saw that day, that time, in that area differ from a car randomly selected from the whole population? Surely the point is that we are aiming to tell that car from any other car that could plausibly have been in the vicinity.
Certainly we should consider to an extent the plausibility that an out-of-town, or even out-of-state, car needs to be included in the population. But we were taking a comparatively simple mathematical exercise, in which the lockedness of the room was taken as a given.
Something is begging to be laughed at, but it’s not the logic.
Dragging in a whole heap of other stuff not in the hypothetical.
Gotcha. So when presented with a mathematical problem including some flavour text, you ignore the mathematical conclusions and nitpick the flavour text. You must have been a joy to teach math to. Your logic is on a par with the pupil in the following dialogue:
Teacher: Suppose X is the length of the hypotenuse…
Pupil: But sir, suppose X is not the length of the hypotenuse?
And the reason why I say this is because even if the witness was able to tell maroons from reds 100% of the time during the “reliability test”, but maintained a 95% proficiency with respect to reds from maroons, the probability that they’d be able to correctly identify the right car from a whole pop of reds and maroons (with 2% maroon prevalence) would still be only 29%. Which is shockingly close to the same number you achieve when sensitivity is 95% (28%).
So despite the fact that the witness was able to correctly spot the “needle in the haystack” 100% of the time under test conditions (and made a few slipups with the reds), according to you they are unreliable. With that kind of logic, no one would take a screening test out of fear that they’d get a false positive. But amazingly, screening tests are used on a routine basis and do a lot of good. Ask yourself why.
Are all cars in the city, state, or country equally likely to be at the scene of the crime exactly when it took place?
Cause that’s what “random” means. If I have a bag full of marbles and pull one of it, I’m assuming that each marble has an equal chance of being grabbed.
The sides of a coin have an equal chance of landing when I flip it.
If a crime has taken place and no one knows for sure who did it, there’s only going to be a select few who are potential suspects. The police isn’t going to think someone who was in the south side of town is a potential suspect in a hit-and-run investigation that took place in the north side. Not all maroon cars that are registered are equally likely to have been involved in a specific hit-and-run incident. If the driver was said to have been male, the list of suspects do not include all the men in the town.
But hey, don’t take my word for it. See “random.” Then tell me if you still don’t see the difference.
That given shouldn’t be taken at face value when you are trying to prove someone is illogical. This discussion is about more than mathematics.
I just had to explain what “random” means to you because apparently you don’t know. Yet you swear up and down that you are right and I’m wrong. I think that’s pretty funny.
Here’s the thing about Zoe’s random number generator which illustrates the error you are making:
In effect, you are looking at a past event and determining that it is unlikely just because maroon colored cars are “rare”. But any event is rare if you seize upon isuperficial characteristics. Such as car color.