Who exactly forms this subset of 1000 then? You're applying the 5% of all men are rapists figure to the 1000 to get 50 men are rapists - does this not suggest that these 1000 men are a random subset which is representative of the male population?
Then you're applying the 2-10%(5% avg) of false accusations to the same 1000 men.
Right, this is the key point that anisotropy has missed in his poorly formulated statistical model. He's taken an analysis of infection rates and applied it to rape allegations, but totally overlooked that that infection and testing are uncorrelated, and rape and alleged rape are not. Anisotropy's analysis is only valid if you randomly allege rape to everyone (in order get lots of false positives of innocent people), just like you would randomly test everyone for infection. The math doesn't work out if only sick people were to get tested.
And then on top of that fundamental misunderstanding, he's artificially deflated the numbers by assuming a "true" incidence of rape that is much lower than the rate of false accusations. This assumption guarantees that most accusations will be false, just by the nature of the problem setup. This has no bearing on reality, even if the allegations and the guilt were truly independent variables. Which as I pointed out in the previous paragraph, they are not.
Hi all,
I can see many still do not grasp the reason why 2-10% should be applied to the (not-rapist) subset. I will focus on answering those that understood it, at least partially.
The math doesn't work out if only sick people were to get tested.This is true, what I formulated here is if we picked anyone at
random without paying any attention whatsoever regarding the specifics of each persons case. Of course it doesnt work if only sick people/actual rapists were to get tested, because that would mean you are dealing with a completely different population composition (rapists only). Statistically it would translate to 0 false positive, which is
not useful nor appropriate when determining the likelihood of someone being actually guilty when we are dealing with a
random person in the population.
does this not suggest that these 1000 men are a random subset which is representative of the male population?What we can do, is we expand the population to include the entire US male population, as long as the composition doesnt change, 5% rapists 95% not-rapists, we will get the same result, try it out. Therefore this criticism is invalid.
artificially deflated the numbers by assuming a "true" incidence of rape that is much lower than the rate of false accusations.my chosen rate of false accusation is 2-10%, avg 5%. What did you mean with "true" incidence of rape? True positives? It is low because 1. rapist population is relatively low, and 2. false negative rate (non-reported incident) being 2/3. This result directly spawns from these realistic/researched rates. You can tweak it if you want.
even if the allegations and the guilt were truly independent variables. |
you are using statistics like who are accused and who makes false accusations are randomOf ALL the criticisms, I find this the
most relevant. Some people are good at statistics here! What you are really asking is does the 2-10% fp rate really apply to the "not rapist" subset, because surely they would be accused less frequently than a real rapist? It is true the 2-10% remain an estimate, past studies put the figure between 0.5% to 90%. Your guess on this is as good as mine. If you can provide studies that shed more light, I would be happy to reformulate the scenario.
Look, we can sit here and quibble if the rates i used (fp, fn, pop comp) are appropriate, but the method i used is sound when we are dealing with a random sample.
We can also sit here and quibble if treating the Kavanaguh case as a random sample (the way i formulated it) is appropriate, but the method is sound.
In the absence of physical evidence, I think what I did here is relevant by seeing it as a random sample.
Finally runbikerun,You say i am intellectually dishonest. But let's look at your bikers doping test example. I do not possess knowledge on how bikers are drug tested, so i went searching on line. This is what i found
https://www.businessinsider.com/how-cyclists-are-drug-tested-2015-9#once-riders-return-from-the-restroom-to-the-doping-control-station-they-sit-back-down-with-the-dco-who-checks-whether-the-urine-sample-is-suitable-10"If the sample is suitable, the DCO tells the rider to select a "urine kit," a box that contains two small glass bottles into which riders will pour their urine. The rider is instructed to check that the bottles are correctly labeled and that there is nothing inside them.
One bottle is labeled "A" and the other "B." Later, after a lab analysis, if the A sample tests positive for a banned substance, the rider has the right to have the B sample tested."This sounds
exactly like what i formulated here: one positive result gives you low certainty, but two positive result boosts it to much higher certainty. And whats more, its super prevalent.
"Teddy Cutler of SportingIntelligence.com recently took a an excellent and detailed look at all the top cyclists from 1998 through 2013 and whether or not they have ever been linked to blood doping or have links to doping or a doctor linked to blood doping.
During this 16-year period, 12 Tour de France races were won by cyclists who were confirmed dopers. In addition, of the 81 different riders who finished in the top-10 of the Tour de France during this period, 65% have been caught doping, admitted to blood doping, or have strong associations to doping and are suspected cheaters."https://www.businessinsider.com/lance-armstrong-doping-tour-de-france-2015-1Your arguments here are thoroughly debunked, doping was found to be common, and the drug test method is similar to what i formulated here. Together with how you thought the conditional rates from one set of statistics would apply to a wholly different scenario (wrongful conviction) your thinking pattern is hilariously original to say the least. I called you banal once, but it's clear to me now you don't lack originality per se, you lack assiduity and cognition.