> -----Original Message----- > From: [EMAIL PROTECTED] [mailto:[EMAIL PROTECTED] On > Behalf Of Peter Flom > Sent: Monday, April 26, 2004 3:38 PM > To: [EMAIL PROTECTED] > Subject: Re: [edstat] 'standards' Was: IQ, Why do we transform
snip.. > One measure of 'success' is income. Here, I have seen studies that > show a curvilinear relationship between IQ and income, largely because > the highest paying jobs (CEO and the like) are usually taken by people > with modestly high IQs (in the 130 - 140 range, IIRC). And the highest > IQ people often enter professions that are NOT that high paying (e.g. > academia and research), but which offer other rewards I'd love to see a reference or two here. No doubt it's not perfectly linear, but I'd be surprised if it's not monotonic. There are a lot of people bright people working in the "private sector" - after all, that's where the money is! I do not have a degree in psychometrics, so I am only speculating, but it seems to me that part of the linear/monotonic issue can be addressed by looking at how IQ changes as we changes a second variable believed to affect IQ. For instance, IQ scores regress toward the mean in a predictable fashion. For instance, the amount of regression to the mean between grandparent and child is twice the regression seen between parent and child. In that sense IQ "behaves" like an interval scale, at least in the middle range. In a similar fashion if we measured a quart of water, at it was 0 degrees C, measured another quart and it was 100 degrees C (boiling); if we mixed them the temperature would register 50 degrees. Similarly if we mixed 3 quart of 0 degree water with 1 quart of 100 degree water, the result would measure 25 degrees. This shows the centigrade scale to have equal intervals because it "acts" that way. My point about IQ being "easier to measure than define" applies to this as well. Ask a high school student to measure temperature with a centigrade scale, and it's done. But define the underlying nature of temperature? It has something to do with Brownian motion of atoms, and this was discovered long after the centigrade scale was developed. > > > Peter L. Flom, PhD > Assistant Director, Statistics and Data Analysis Core > Center for Drug Use and HIV Research > National Development and Research Institutes > 71 W. 23rd St > www.peterflom.com > New York, NY 10010 > (212) 845-4485 (voice) > (917) 438-0894 (fax) > > > > >>> [EMAIL PROTECTED] 4/26/2004 3:18:22 PM >>> > > There is a tremendous amount of scientific literature on the > measurement > of intelligence, much of it unknown outside the discipline, even > amongst > psychologists from other specialties. Gardner's multiple intelligences > is not held in the highest esteem in the field. You don't see well > developed operational definitions with high reliabilities and > validities > for his various intelligences (which might be better described as > talents or abilities). > > On the g issue, correlations between real-world success and > "intelligence" is highest when the test is heavily loaded on g (e.g. > Raven test). Contrary to what a previous person said about the Raven > test, one does not memorize the answers (there is nothing to > memorize). > Some "g" items look like they might involve memorized material, > actually > do not. This is an arithmetic reasoning item (not a Raven item) that > would contribute to g: "Bob is twice as old as his sister, who is now > 7. > How old will Bob be when his sister is 40?" The "math" is not the > issue > - it's the ability to represent the elements symbolically and > co-ordinate the different parts of the problem. People more "heavily > load" with g can do these faster and more efficiently. Though in > fairness to critics, we can measure intelligence better than we can > define it (i.e. we can show it "by site if not by nature").Tests that > utilize more rote memory are lower and g, and, show worse psychometric > properties, such as validity coefficients. The fact that factor > analysis > was involved in demonstrating g lead many people to view it as a > statistical abstraction only, but the collective evidence suggests > this > is not so. > > For years people used to claim g did not correlate that well with > "real > world" success, but studies with large N show a remarkable effect of > g. > Ree & Earles (1990) "Differential validity of a differential aptitude > test" AFHRL-TR-89-59 (Brooks AF Base, TX) reports % of "training > success" in different military specialties explainable by g (I round > to > nearest %). > Nuclear weapons specialist 77% (i.e. rocket scientist!) > Weather specialist 70% > Fireman 60% > Vehicle maintenance 49% > Maintenance 28% [to mention a few] > > Hunter and Hunter's meta-analysis (1984 Psych Bulletin) across a wide > variety of professional occupations finds cognitive test score (i.e. > IQ) > the best overall predictor (r = .53); second is biographical data r = > .37, and college grades is a feeble .11! Interestingly, twin studies > find that of all the major components of intelligence, g is the most > heritable. Psychometrics is an interesting aspect of psychology. The > "controversies" regarding these tests _outside_ the field belies the > consensus within the discipline regarding the definition and > measurement > of intelligence. > > ============================================ > John W. Kulig > Professor of Psychology > Plymouth State College > Plymouth NH 03264 > ============================================ > > > . > . > ================================================================= > Instructions for joining and leaving this list, remarks about the > problem of INAPPROPRIATE MESSAGES, and archives are available at: > . http://jse.stat.ncsu.edu/ . > ================================================================= ============================================ John W. Kulig Professor of Psychology Plymouth State College Plymouth NH 03264 ============================================ . . ================================================================= Instructions for joining and leaving this list, remarks about the problem of INAPPROPRIATE MESSAGES, and archives are available at: . http://jse.stat.ncsu.edu/ . =================================================================
