Re:Women and Minorities in Science (Warning: Basic calculus ...
by _h4xr on Tuesday October 09, 2007 @04:57PM
(#65461)
I don't think they're beyond question. Certainly, there's room for improvement (the cutoffs are too low. A physics PhD program isn't really interested in a test that has no measurement value for IQs over 120, or wherever it is -- and the analogy section was pretty g-loaded, so it's a shame that's been removed from the SAT, too). And schools don't just rely on standardized tests; they'll look at outside achievements, school grades, and maybe essays -- a competitive program might want people in the top 5% of intelligence, determination, and charisma -- which still means their students' academic test scores will need to be in the top 5%.
2) That holds true for studying results within schools, but not for test results overall. This is because schools admit people with roughly the same ability -- so they might have one person with test scores .5 stdev below the mean, but with a really brilliant essay. Fortunately, you don't have to take my word for it -- just look at the dropout rate among people who are admitted despite lower test scores and lower grades: http://www.capmag.com/article.asp?id=1635
2) Nothing. I just doubt that schools have actually been magically fooled into thinking they do: They actually get to do the regression analysis, and see that past grades are a better predictor than standardized tests. This has been well known for a long time. If applicants have been fooled, however, it allows the school to get useful effects from the test (even if the effect isn't predicting future performance).