Comparative Error Rates

5 Comments

1. 

   CStanley

   November 7, 2006 | 7:39 am

   Where to start…

   Let’s see, I guess I’ll address your points in the order that you wrote them.

   My first observation is that Mr. Crittenden seems utterly unconcerned with the potential for Type II error. That is, he seems quite confident that all or most accusations of racism are false positives, and that it is highly unlikely that they have any grounding in reality

   Where in the hell in anything Crittenden wrote can you conclude that he is “utterly unconcerned with the potential for Type II error”? Because he didn’t write about this concern in this particular article? So, anytime you write about one concern that you have, you automatically are making a statement that you are concerned about nothing else? Or is it that you feel it must be mutually exclusive to be concerned about what you call Type I error and Type II error? There’s a false dichotomy for you.

   In fact, I’d say that your insistence that it must be one or the other is the problem. You see, if people are truly concerned about racism and its harmful effects, one must be concerned about both types of error. But those who make the claim, as you do, that Type II errors are all important, that we must accept overwhelming false positives, are the ones who are guilty of not having concerns about the real effects of both types of errors. When medical diagnostic tests are designed, it is extremely important to limit both types of error; otherwise grave treatment errors will result. Treating a disease that doesn’t actually exist in the patient can do as much harm as not treating a disease that does exist. In fact, sometimes more harm, because the body often has the ability to heal itself and as physicians acknowledge when they take the oath to “do no harm”, the cure can be worse than the disease. The parallel with racism is that to some degree, our society must heal itself, not have a cure forced on it by the thought police. Racial wounds will take time to heal, but constantly picking the scab will only slow that process.

   Now, are some liberals sincere and well-intentioned in their belief that pointing out all possible instances of racism is helpful (thus promoting the idea that it’s better to accept a high degree of false positives)? I’m sure that many are. But is there also a more sinister motive in the modern Democratic party? I believe so, and it is this: that keeping the specter of racism among the GOP alive is an easy scapegoat to keep blacks beholden to their party. It’s no secret that many blacks are becoming increasingly disenchanted with the Democratic party and feeling that they’ve been taken for granted, and as that occurs, I think we can expect more and more instances of the race card being played by the left.

   Now, on to your comparison of racial accusations and disloyalty accusations.

   And like with race, when the accusation is leveled without cause, it should give us pause–false claims of disloyalty are serious offenses that undermine the political process and pervert democratic legitimacy. So if we are more likely to make false IDSs of “disloyal” criticisms compared to false IDs of “race” criticisms, then we should be more concerned with candidacies that seem likely to exacerbate this pre-existing bias.

   OK, IF we are more likely to make the false ID’s of disloyalty criticisms. That’s a big IF, and not an easy one to prove.
   Not that you haven’t tried:

   A quick look at history shows that accusations of disloyalty rarely actually pan out. Across the Red Scare, Japanese Internment and McCarthyism (to name a few), the trend has generally been that we’ve accused people of being disloyal when there was no grounds for it (Type I errors). There’s no reason to think differently of the modern day–for all the talk about Democratic disloyalty, Nancy Pelosi has not been caught spying for al-Qaeda and Baghdad Bob was not Harry Reid in disguise.

   To get back to your comparison, this would be like me saying that the GOP isn’t sponsoring lynchings of black people, so therefore we can assume that all accusations of racist intent by the GOP are false. That you don’t see the double standard you are using is a clear indication that you were biased to prove your predetermined conclusion. If you would like to consider real acts of disloyalty, how about Ted Kennedy cozying up to the KGB during the early ’80s in order to make political hay against presidents from both parties, how about Chinagate, how about the fact that Kerry’s explanation that what he meant to do was call the POTUS an idiot is now considered completely acceptable as a form of responsible dissent?

2. 

   Kim Ritter

   November 7, 2006 | 2:14 pm

   gee- CS I thought you were above the Karl Rove style mud-slinging. Isn’t this exactly what you called me on the other day???

3. 

   CStanley

   November 7, 2006 | 2:17 pm

   Mudslinging? What are you talking about?

4. 

   David Schraub

   November 7, 2006 | 3:47 pm

   Calling a President stupid is not “disloyal” under any metric I can think of (though labeling it as such is game, set, and match for my claim that we’re very likely to make Type I errors with regard to disloyalty). Moreover, you miss the point of how Type I and Type II errors interrelate. Obviously, its good to reduce both. But as a general rule, when you put more effort into reducing one, the other goes up. This is axiomatic–by making it harder to disprove the null, you reduce the likelihood of Type I errors but simultanously raise the likelihood of Type II errors. If you make it easier to disprove the null, the reverse happens. I’m saying that empirically, we tend to err on the side of Type I errors with regard to disloyalty, and err on the Type II errors with regard to race. If we want to balance that out, we should be more concerned with the prospect of uncaught racism than we should over racism “false positives.” And again, given the overwhelming data showing continued Black disadvantage in American society (economically, politically, and socially), I think it’s facile to say that racism only remains as an issue to demonize the GOP (though that again assumes that the error is Type II rather than Type I, biting into my critique even harder).

5. 

   C Stanley

   November 7, 2006 | 11:06 pm

   Moreover, you miss the point of how Type I and Type II errors interrelate.

   No, I’m quite aware of the relationship between them. What I reject is the premise that we must accept wide variance in either one, or that people who express concern about one type of error are necessarily unconcerned with the other.

   To again use my analogy of medical diagnostic tests: frequently such testing is done with a two tier approach. Screening tests are designed to accept greater Type I error or false positives. But then a second test (which is more inclined to give Type II error and thus would have missed some true positives) is used to limit the error by either confirming the positive or giving data that may lead to rejection of the positive result. However, in the case of evaluating claims of racism, there is no second test, and accepting the high rate of false positives as you propose (and as many liberals currently do) has the effect of levelling false charges which cannot be disproved. And, my argument is that this should be avoided because it actually exacerbates racial division in our society instead of helping to eliminate it.

   You are correct in the argument that the relative importance we place on each source of error should relate to the actual, true prevalance of the problem that we are trying to predict. But the relative importance we place on these errors should also be dependent on the actual outcome which results from accepting either false positive or false negative results. Which of these situations is more harmful? If you aren’t asking that question, then you are working the wrong problem.