Wednesday, November 4, 2015

Three Indirect Measures of Important Things

In this and the next postings I'm going to identify three important things that can be known about oneself which can be measured indirectly. If measured directly, there can be considerable hassle, as in the case of measuring one's O2 Max, a key fitness indicator, or in the case of REM sleep, where EEG leads need to be placed on your scalp.

Some might wonder about a direct measurement of intelligence. There is probably no "magic bullet" to do this. That's why we have intelligence tests and even aptitude tests which purport to give some indication of brain power. But what if there were a way to predict performance on an aptitude test, say the verbal part of the Scholastic Aptitude Test (SAT) not by answering questions about difficult paragraphs, but on the basis of what you prefer; in effect, on the basis of a preference test. It would seem no cognitive test is being given (nor is it), yet this preference test can predict results on aptitude tests.

That's what triggered an interest, many years ago when I was a college freshman at the University of Michigan. I had just answered a series of preference questions (Agree or disagree: "I like cooked carrots rather than raw carrots," or "My stools are black and tarry.") These questions went on seemingly forever (actually, there were only 396 or them), but on the basis of these preferences, it was possible to predict my verbal SAT score almost exactly. I suspect that they could do the same with the math part as well. And maybe they did, but I to remember the verbal part. I was quite intrigued by this.

I had just taken what was affectionately known as Benno G. Fricke"s "Raw Carrots" test, which seems magical, indeed, because it could predict skills and aptitudes which would seem to have nothing to do with the chosen preferences. (I notice that Dr. Fricke was defending his 396 question test as late at 1975 in this Michigan Daily article.)

Just how this feat was accomplished I leave to colleagues far more versed in statistics than I am. But the basic idea is simple enough: you run correlations between student responses to the preference questions and what the same student got on his SAT test. (Michigan had all of this data, and the willingness to crunch it before computers went big league.)

Patterns would emerge. Perhaps on a subset of the preference questions. It might have worked this way: a set of responses to the 396 questions would emerge that correlated with top performance on the skill or aptitude test; another set with a less stellar performance, and so on. Within the grouping (all college freshmen), stable correspondences could be found. The test would not work, for example, on a random set of the US populations - I would think.)

So within this population, the predictions about aptitude, based simple on preference, would work and would seem magical.

The utility of the preference test does not matter. Some might say, you have the aptitude test results. What do you need the "predictor" tests for?  Good question. What if someone didn't have a SAT score. No problem. Fill it in with a really good estimate. Or perhaps, just perhaps, it would be interesting to see whether the preference test could predict skills in artistic perception or spatial perception, say in Gardener's 16 "intelligences. But this is for another time. For the present, note the ability to predict SAT scores indirectly.)

Okay. Enough about raw carrots. Now to the first of two measures that are important in assessing health: the indirect O2Max measure and the indirect REM measure.

No comments:

Post a Comment