« Day 10: Most secret origins usually don't involve tequila | Main | Bali, plus ten »

November 11, 2007

Comments

Feed You can follow this conversation by subscribing to the comment feed for this post.

Noel Maurer

I'm a "liberal airhead." I love that!

Of course, the questions in that survey are loaded. Frex, the three about disciplining children, using former Nazis in the administration of post-war Germany, and the value of military service are all impossible for a liberal airhead to answer honestly due to the way they're phrased. I tried, which kept me from being a "whining rotter," but actually seeing the survey up-close hasn't done a whole lot to change my previous opinion of Adorno.

Judging from your post --- and yes, I am aware that I'm taking this way too seriously, but the Green Bay game is about to begin so why not --- we still seem to be pretty far away from even a good micro-theory of the wingnut, let alone a grand unified one. All we've got are some intriguing results from data-mining and a vague set of if-then statements. Right?

claudia

I mentioned the problem with the engineer's association in Jordan a few months back, right?

The F-test is sixty frickin' years old. We should have something much /much/ better by now.

The homophobia paper seems a bit like testing for whether small boys like things that explode. On the other hand, it's good to get objective verification for these things. What was the correlation?

Better discriminator, forsooth. Hey, Mavissikalian is an Armenian name.


Doug M.

King-Walters

On a related issue, how trustworthy is Altermeyers work on this subject? If it is reliable, it could be relevent.

Carlos

Noel, I'm shocked that you think Cohen's analysis of the 'Stache is anything but serious.

I think we're in the position of linguists in 1700 with collections of word lists, seeing resemblances between languages and sensing a deep and pervasive pattern. The tower of Babel? The sons of Noah? Aeneas fleeing from Troy? I'd like to think that modern statistics helps us winnow the wheat from the chaff a little better. I'd really like to think that.

Doug, it's analysis of variance, not linear regression. The variance caused by the two groups (homophobes and decent people) considered as units was large compared to the variance within the two groups, at p < 0.001. Here's the breakdown: "In the homophobic group, 20% showed no significant tumescence, 26% showed moderate tumescence, and 54% showed definited tumescence to the homosexual video; the corresponding percentages in the nonhomophobic group were 66%, 10%, and 24%, respectively." Before you ask, 64 white male college students at the University of Georgia, 29 decent, average age 20.

The graphs comparing tumescence over time between groups for each video is kind of funny too. But they all got partial course credit, so.

The cite, incidentally, is misspelled in the original: it should be Mavissakalian. "Responses to complex erotic stimuli in homosexual and heterosexual males," British Journal of Psychiatry, March 1975, 126:252-7. I think it's the same guy currently at Case Western.

KW, I've kept on meaning to read Altemeyer in depth. But I've only paged through.

Noel Maurer

Data mining can always be useful, but it's still a problem when you don't have a theory. Consider the title of the paper: "Is homophobia associated with homosexual arousal?" Says nothing about causality. Could there be reverse causation? I could come up with a story for that. Some common third factor? Ditto. Who knows?

Carlos

Actually, *that* paper does eliminate one theory: homophobia is not linked to aggressive personality as measured by standard questionnaire.

It doesn't distinguish between the two other main competing theories, latency -- my personal favorite, but extremely difficult to test -- or anxiety, which has been documented as increasing arousal and erection within this sort of protocol.

There's also a subsidiary result: homophobes report themselves as being less aroused than what the MIR indicates, but all other combinations show a fairly accurate self-assessment, and response discordance *is* a known psycho[logical, not analytic] variable.

So. Not simple data mining. Psychometrics is pretty different from econometrics, and the stuff I learned in the wet sciences is different from both of them (though closer to the former).

[rummages] Jesus, Kennedy only has a page and a half on ANOVA. WTF? "In light of the above, it can be concluded that anyone comfortable with regression analysis and dummy variables can eschew analysis of variance and covariance techniques." Uh-huh.

Michael

People, people -- Bob Altemeyer at the U. of Manitoba is the go-to guy for the sixty years of research that build on that quiz.

His entire book-for-the-layman is online at http://home.cc.umanitoba.ca/~altemey/ and yes, it's damned good.

Michael

BTW, I appear to be a whining rotter, at 1.67. I tried to answer honestly, really I did.

Carlos

... um. If I had read that first, I might think he was a well-meaning crank.

Since two people have mentioned him, here's my thoughts on the Altemeyer I have read. Altemeyer has shown a distinct RWA group exists. (Which we kinda guessed already.) But he's short on mechanism, and his results mainly link score to score, not score to behavior.

To bring it back to the original paper mentioned, what's the causal connection between choosing a career in engineering and scoring high on an authoritarian survey? It's not obvious.

Noel Maurer

On homosexuality and homophobia: fair enough, Carlos. The study does eliminate one hypothesis, and that is indeed progress. I'm not sure what you mean, though, when you say that psychometrics is different from econometrics.

On the F-test and engineers: one possible hypothesis is that the test is actually measuring a relative lack of social skills among engineers, rather than a relative tendency towards authoritarian impulses.

Hark back to what I said about the F-test questions being loaded. Anyone who has internalized the prevailing social norms, /even if they have authoritarian tendencies in fact/, will score lower on that test than their "true" authoritarian tendency.

Frex, when I said that I answered three questions "honestly," what I meant was that I actually clicked off a /more right-wing/ answer than my initial impulse. In at least two cases, I had to swallow my bile to do so. In other words, I had to answer three questions "dishonestly," by surpressing my entirely valid emotional reaction to the wording. Otherwise I'd be a "rotter" too.

The only way to test the "social skills" hypothesis would be to develop a more-accurate and less-culturally-loaded authoritarianism test, because this one is going to miss a number of actual right-wingers who have emotionally-internalized the prevailing social norms. Such a test might find no difference between engineers and others; the "crazy engineer" phenom could be simply a small but highly visible right-tail.

Or not. (I suspect not.) But that test is loaded, and it is going to miss a lot of right-wingers.

Bernard Guerrero

"But that test is loaded, and it is going to miss a lot of right-wingers."

I'll say, only a 3.13, and I'd oppress the lot of you on a dime.

Andrew R.

I am always a bit suspicious about any methodology that involves a man having instruments attached to his bits in a laboratory setting. Sexual arousal being somewhat contingent on a whole raft of circumstances, the results you get from surveys like the one mentioned seem like they'd be very much in the "extremely broad and not terribly useful" range.

I seem to remember that the methodology was similar for the test in which some researchers came to the conclusion that there is no such thing as male bisexuality.

Carlos

Psychometrics developed from the experimental tradition of statistics. It grew out of the need to create reliable methods of psychological measurement. Econometrics came from the observational tradition of statistics. It grew out of the need to test economic hypotheses against sketchy data where experiment was impossible.

They're very different; and both are different from the basic experimental statistics I learned in college and grad school, which had the luxury of both controlled experiments and a strong microtheory.

Adorno's test is old, and I'm not going to defend it as anything more than a parlor game. (You can talk to Daniel Davies about that.) But your *type* of response -- it doesn't matter in a well-designed test. Which is creepy.

The questions then become, what is the test actually measuring at the microtheory level? and what are the behavioral correlates? Which is where I think Altemeyer has some gaps. Cosma Shalizi somewhere half-wishes that RWA be used as a factor in a repeat of the Milgram experiments, which is just not going to happen.

Carlos

Andrew, how can you say that was a laboratory setting? There was a comfortable reclining chair!

I'll clip and post the graphs, which are hilarious.

Dennis Brennan

quoth Carlos: "I think we're in the position of linguists in 1700 with collections of word lists, seeing resemblances between languages and sensing a deep and pervasive pattern. The tower of Babel? The sons of Noah? Aeneas fleeing from Troy? "

Is this supposed to be a description of wingnuttery or of wingnutology?

Michael

Noel says, "The test is actually measuring a relative lack of social skills among engineers, rather than a relative tendency towards authoritarian impulses."

Yeah. I'd buy that.

Carlos

Dennis: wingnutology. I chose that example for a reason. Classical comparative historical linguistics is the most rigorous social science I know of. Predictive, Popperian. When biologists use their computerized cladistic models on linguistic data, they get garbage results, ahistorical nonsense. So its methodology actually ranks above some of the hard sciences.

In 1700, however, you had a lot of groping and crazy-sounding speculation.

(Since the subject is a crank magnet, you still do; and since a large fraction of the theoretically attainable results have been attained, it's becoming an orphaned methodology, with serious researchers trying to push the edges by accepting less rigorous methods. I suppose it's parallel to what's going on in particle physics.)

Cosma

"somewhere half-wishes that RWA be used as a factor in a repeat of the Milgram experiments, which is just not going to happen."

Actually, Altemeyer says he did just this study in the early 1970s; I haven't tracked down the reference to see if he did it well. It's certainly not going to be replicated, however.

The comments to this entry are closed.