Saturday, April 01, 2006

GRE and IQ

I was planning to write a full response to this blog entry by Prof. Motl, about IQs as deduced by the GRE scores of applicants to various doctoral fields. But who cares? You can find and download Powerprep from the GRE website and examine the data for yourself. You'll need a Microsoft Windows PC.

I'll simply post two charts here, and note that if GRE scores are linearly related to IQ (as presupposed in the link) then the candidates for graduate school are not normally distributed in IQ. But it should be clear that GRE quantitative scores do not serve well to distinguish between candidates at the high end of the scores.

gre_verbal_all

gre_quant_all
at

4 comments:

  1. Luboš Motl3:52 PM

    Dear Arun,

    of course that college students don't follow a normal distribution of IQ.

    Only the full population can be expected to exhibit the normal distribution.

    The college students have been selected by a special procedure. Essentially, you may imagine that their IQ had to be above 95 or so when they were admitted to the college.

    This means that the IQ distribution of college students will have a rather sharp lower end, which is exactly what the graphs show. If you think for 30 seconds, it must be completely obvious that this is exactly what is expected.

    I don't want to claim that the relation between the IQ and the GRE score is totally correlated and exactly linear, but your asymmetric curve *strengthens*, not weakens, such a hypothesis.

    Best
    Lubos

    ReplyDelete
  2. Arun5:24 PM

    As I said, take a look at the data. E.g, the verbal scores for engineering students are in a bimodal distribution. The perfect quant. score for the entire population is at something like the 90th percentile; among engineering and mathematical sciences 800 is at something like the 75th percentile.

    ReplyDelete
  3. Luboš Motl11:49 AM

    Dear Arun,

    debates with you are completely irrational because you're unwilling to think even about completely elementary points. What is your last comment supposed to mean? That different groups have different distributions for different abilities? What a discovery.

    Best
    Lubos

    ReplyDelete
  4. Anonymous2:25 PM

    "of course that college students don't follow a normal distribution of IQ."

    Lubos: your English Grammar is still very slightly imperfect (unlike your string theories)

    ;-)

    ReplyDelete