Yes, but to draw that conclusion he had to add assumptions to the hypothetical. In the hypothetical, if the group in question was 99% white and 1% black, so that rape rate and population rates were in line, his argument wouldn’t fly. He had to skew the populations out of synch with the rape rate to reach his conclusions. The problem with attempting to apply Bayesian inference to this is that it requires more parameters than we gave in the hypothetical. He chose to add his own to make up the difference.
Thanks for you reasoned contribution to this thread, asshole. We are so astounded by your knowledge that we hardly know what to say.
I don’t see how you don’t recognize a difference between saying “25% of all rape accusations are false”, and then interpretating that to mean that both groups contribute an equal number of false accusations.
If he wanted to reach the conclusion that he did, the given would have to read “black women and white women lie about rape 25% of the time”.
And his conclusion is crazy because he is saying that a black woman has a greater chance of lying than a white woman. What he really means to say is that if you were to randomly select a report of wob rape, the chances of this claim being false are greater than if you were do the same experiment with bow rape. But big deal. This is no difference than predicting a dice roll.
An event has already happened and you have nothing to go by except the claimant’s word. If you hold that both groups of women are equally honest, then they are equally credible. Reverse probabilities do not apply when the dice has been rolled and you’re asking someone to tell you the outcome.
Gorsnak, can you please tell me if this makes sense to you? I don’t want to see you go down defending an illogical conclusion, but that’s what it seems like you’re doing.
I’m pretty much done with this thread, but this caught my eye as an opportunity for one of us to see the light. The difference is that a roll of a die is completely random. You have a 16.7% chance of it coming up any of the six possibilities. If the die, instead had three red sides and three blue sides, we’d have a 50% chance of it comiing up either red or blue. But that does not represent the question. We’d have to have a die that has four red squares and two blue squares. There is one assumption in this, and that i9s the incidence of the two types of rape will not change from what it has been. I think that is a fair assumption because anything not stated in a hypothetical is assumed to remain constant.
So if this die was rolled every day for the past ten years and resluted in it coming up red 67% of the time, we can expect it to come up red 67% of the time in the future. It’s the same with the rapes, we can assume that b-o-w and w-o-b (or w-o-w and w-o-b, which I think was the original scenario) would continue at the same proportion.
So whatever roll I chose to look at, both in the past and in the future, will have a greater likilehood of having a red (non-blue) outcome. Similarly, whatever claim of rape I look at in the past, or in the future, will have, a greater likelihood of having a white victim.
Ooooh, I hope so. 
Okay, I’ll accept a die with 4 red sides and 2 blue. I’m interested in how that changes things.
Yes. But remember that hypothetical I gave you where I asked you to tell me the outcome of a dice roll? After the die has been rolled and the outcome has been observed, prior prediction has no relevance at all. Before we roll the dice we predict that 67% of the time it shall came up red. But when the roll has been made and the outcome observed, none of that matters anymore. We know that it is fully plausible that a blue has come up, so the question is no longer “what is the more likely event”. The question is “what is the chances that the person reporting the event is lying”.
Just because more white victims are expected in the future doesn’t mean that when black victims occur, their claims are less than fully plausible. By saying that you are more apt to believe reports of w-o-w rape than w-o-b, that’s like saying you have a hard time believing that your die could come up blue on a single roll.
Tell me, if we rolled such a die, would you be skeptical if I told you it came up blue? Even if I am honest as the Zombie Ghost of George Washington?
Because the women who file a false claim about being raped weren’t raped. The assumption being made is that the number of false claims will be proportional to the total number of women, not the number of women who were raped. Women who are making false claims are not “lying about rapes” because there WAS NO RAPE. They’re lying about “not-rapes”. So why should we expect their numbers to be proportional to the number of rapes, rather than the number of “not-rapes”?
And again, just so there’s no misunderstanding, I don’t expect this assumption to hold in reality, because I don’t think that false claims of rape just spring up out of nowhere. But jeez, within the stated framework, what the heck are you having such a hard time with?
And what I’m getting at is that the 25% statistic applies to claims only. Not to people. It’s a rate based on total accusations made. To infer that it applies to people is to completely disregard what we are told.
He is assuming a 1 to 1 stoichiometry of liars in both subpopulations. But again, we know nothing about the people in both populations. We only know about their claims. To assume that we know stuff about the people is to introduce information that wasn’t stated upfront.
Because we are told that 25% of all accusations are false. We are NOT told that 25% of the people in both populations lie about being raped. See the difference?
We’re not saying you believe in that assumption. Me, personally, have a hard time understanding why you think his answer follows from the supplied givens. I’m just trying to show you why I think he’s wrong in a logical sense, not necessarily mathematically.
Except if the die has been rolled enough times to establish a definitive bias. Our die with four red squares has a bias built-in. A history of prior rolls would only confirm that. With our rape victims, the question presents a similar case of non-even distribution as a fact—there are more w-o-w rapes than w-o-b rapes. A fact that we have no reason to expect will change in future rapes, as none of the conditions supplied implied such.
It is true that prior rolls do not effect future rolls. But ALL rolls have a bias, the same one: that there is a 67% chance that it will be red. So any isolated roll, whether in the past or int the future, is more likely—67% vs 33%—to have a red outcome. Similarly, if we choose any rape (forget claims of rape for now) and look at it, there is a greater likelihood that we will find a white victim. I think we’d agree to that point. Correct?
So, now we add the fact that claims of rape have been made, one by a black woman and one by a white woman. We are told that only one is telling the truth, but as far as we know they are both equally credible as people. But we know that one claim is not credible. Can we glean which one is more likely to be making a truthful claim? Or do we have to throw up our hands?
I maintain that if we have to choose (represented by my cash inducement), that we can make a decision that has a better probability than 50-50 of being correct. Since we have no reason to believe that the incidence of rape has changed or its distribution, there are more actual white rape victims than black ones. A claim of rape, for it to be truthful, has to align with one of those actual rapes. And since there are more actual rapes, there is a greater chance that the claim of the white woman is truthful.
Keep in mind, that the conditions stated that on of them was lying.
No, they are both perfectly possible outcomes. But if you hold a gun to my head and tell me I must bet my house on one outcome or the other, I’d have to bet red. Wouldn’t you? Either way it is still a gamble (either woman might be telling the truth) and I would avoid the gamble in the real world (I wouldn’t want to risk my house or not give both women the benefit of the doubt), but the hypothetical asks us to step out of the real world.
No, the outcome of any isolated roll would not make me skeptical. But that is not the question. The question is artificial in that it forces you to make a decision, but that is what it does. If the ghost of George Washington told me it was blue and the ghost of Abe Lincoln told me it was red, and I was told that one of them is lying, barring any other info, I’d have to be more skeptical of George’s claim.
Right?
So what? We aren’t talking about predicting who will be the next person to get raped by a white male. We are talking about who do you believe when they say they got raped by a white male.
This statement causes androids to self-destruct, just so you know. If we know that one of them is lying, then we also know that one of them lacks credibility. They can not be equally “credible as people”. If we can’t agree to this, then there is no use in continuing.
Good. I’m now going to pretend that we agree with one another on everything else, because essentially what I’m saying is if you are skeptical of w-o-b rape accusations because they are “uncommon” that’s as stupid as being skeptical that a die turns up blue.
Using that logic, since there are more raped white women, there are correspondingly more white women lying about being raped, so no, the white women is not more likely to be truthful. Using the 25% number, 1 out of 4 white women who claim to be raped are lying and 1 out of 4 black women who claim to be raped are lying. Both are equally likely to be lying or telling the truth.
We can move on, and probably should, buty I think this is the crux of the problem. I’d say that prior to the two women making claims we have to view them as being equally truthful. After they make the claim and and before we know for sure which is lying, we also have to assume that they are both equally truthful. But we have another piece of information: there are more w-o-w rapes than w-o-b rapes. So while we have to assume the the credibility of both woman is equal, the claims themselves have different levels of credibility attached to them.
Let’s change this slightly. One man makes two claims. He says:
On Tuesday a white woman was raped by a white man.
At Wednesday a black woman was raped by a white man.
Again, we say that only one statement is true. Would you still maintian that both statements are equally credible? That they have an equal likelihood of being true?
I thiink this change helps things as it removes part of the emotional component that makes all of us uncomfortable, as it is difficult to separate the credibility of the women themselves with the inherent credibility of their respective statements. Here we eliminate that problem completely.
For this to hold true, we’d have to assume that people lie about interracial rape at different rates than they lie about rape by their own race. I haven’t seen a single cite that even obliquely makes that suggestion. Rape is rape, and a given black women in the state of North Carolina is much more likely to be raped than a given white women. Do we trust the white woman less because of this?
Most of us would say no, but what’s your take?
I think the problem is the insistence that being skeptical of the claim is the same as being skeptical of the person. I don’t think it is. What is your take on the hypothetical I posed in Post #351 above? It eliminates that problem to a large degree.
It doesn’t even address the problem. To address the problem of credibility in your scenario, we’d need just two pieces of information. What percentage of white women claiming to be raped by white men are lying and what percentage of black women claiming to be raped by white men are lying. With those two pieces of information, your question can be answered and absolutely no knowledge of frequency of occurrence are needed. That’s all any of us have been saying all along. Frequency of occurrence of a plausible event is not a measure of credibility in any way, shape, or form. We even showed you many real world examples that demonstrate this. Frequency of lying about an occurrence can be applied as a measure of credibility, if one chooses, even though I’m still a fan of believe until you have a valid reason not to.
Yes, it would be great to have those two pieces of information. But that paints a different hypothetical. One that I agree that it would lead to a better answer. But the question is without it, are we forc ed to simply throw our hands up in the air? Or can we make a better guess and a worse guess? I say we can.
Could you please answer the questions I asked in the hypothetical in Post #351. I’m curious as to what you would answer and why.
magellan, answer me this.
Scenario one:
The odds of winning the lotto are 1 in a million tickets sold.
The winning numbers are drawn and a winner allegedly is a New Brunswick retiree who lives in an RV with his wife and Jack Russell Terrier.
Scenario two:
The odds of winning a door prize at a company party is 1 in 50.
The winning number is drawn and a winner is allegedly the boss’s secretary, a sassy blonde who surfs the web on her lunch break.
If someone held a gun to your head and told you to pick which scenario is more apt to be true, which would you pick and why?
I assume there is another condition: that the lottery and the office drawing were actually won be someone. If so, I’d say that both claims are equally likely. I cannot judge any further. If thae gun was to my head I’d pick the boss’s secretary, just for the hell of it.
But this is not a good analogy, as they are two completely independ acts with no relation to each other whatsoever. In the rape hypothetical we have more information pertinent to both and tying them together.
If your thinking is correct, please explain to me how you would answer the questions to my hypothetical in post #351. It is your hypothetical with one noise variable eliminated.
No, we don’t. The rape hypotheticals exist independent of one another. The likelihood that Janice was raped has no influence on the likelihood that Marie was raped.
Janice and Marie are exactly like the hypothetical I presented. The odds that Janice White will be raped are higher than Marie Black’s, just like the odds of winning a door prize are higher than winning the lotto. Yet when they claim to have been raped, both are equally credible in the absence of any other information.
The man making the claims knows which of the situations is false. So the question of concern, again, is not which situation is more likely. It’s which situation would he be more apt to lie about. And we don’t know that. So it’s a blind guess.
If you think about it, the hypothetical you just painted is sorta like mine. I made two claims about a lotto and a door prize winner, and you chose to view them as equally credible in the absence of other info. Your hypothetical claimant made two claims about a white and black rape victim, and I choose to view them as equally credible in the absence of other info.
As I’ve stated repeatedly, I would find them both equally credible, because both are perfectly plausible, and because we don’t have any data that says that one is more likely to lie than the other.
As a side note, in the corresponding GD thread, it was shown that Bayesian Inference would actually consider a white woman claiming rape somewhat less credible than a black woman in the state of North Carolina. I don’t agree with that either, but I thought it might be a data-point you were interested in.
The only circumstances in which using any of the various given theories would result in the black woman being less credible are if black women lie more frequently about interracial rape than rape as a whole. Not a single cite has been given to even attempt to support that theory.
But we have info linking the two claims, i.e., that there are more w-o-w rapes than w-o-b rapes.
To be accurate, that was not your hypothetical. In your hypothetical, neither person made any claim. Both claims were made be a third party who was merely relaying information that each game of chance was one and described the winner. Just wanted to keep things straight. That said, I agree with this statement, but it hinges on the “absence of any other information.” But we do have other information: that thereare more w-o-w rapes than w-o-b rapes. Why do you keep ignoring this piece of info? Do you really believe it is of zero value?
This is going to get my head to explode, and I’m not a droid. It might be a guess, but it is an educated guess, not a blind one.
Scenario 1
Two men each make a claim, one of them is lying.
One man say he’s been to Columbus, OH, the other says he has been to Cincinatti, OH.
Who is more likely to be lying? We can’t tell. Right.
Scenario 2
Two men each make a claim, one of them is lying.
One man says that he’s been to Colombus, OH, the other says he has been to the moon.
Who is more likely to be lying? The guy who claims to have been to the moon, of course.
Now it could be that we are an astronauts convention and one of the guys is Neil Armstrong. But if we know nothing about the people, or their propensity for truth telling, AND we have information about the likelihood of the event happening regardlerss of the people involved, then we can make a better guess and a worse guess.
Do you really think that Scenario 1 & 2 are equivalent from a probablistic standpoint?