I'm going to link back to an earlier post of mine:
http://giftedissues.davidsongifted....sion_to_the_Mean_Why_tes.html#Post230897regarding regression to the mean and repeated testing. Although the main discussion point was how scores fall over time, one may also look at it from the angle simply of, if you test enough times, eventually you will get an unusually high result, that isn"t statistically representative of the bulk of the test scores, but has a better chance of meeting your cutoff criteria. So if the district allows multiple entry years to the same pool of children, and especially if the general population is just outside of the cutoff range, testing every child who has not previously qualified, while keeping every child who has in the program, means that one will accumulate many more than the nominal expected percent of the school in GT.
Quote:
"The “or” is not defensible, however, when both tests are assumed to measure the same construct. For example, the test scores may represent multiple administrations of the same ability test or consecutive administrations of several different ability tests. Error of measurement is defined as the difference between a particular test score for an individual and the hypothetical mean test score for that individual that would be obtained if many parallel forms of the test could be administered. The highest score in a set of presumably parallel scores is actually the most error- encumbered score in that set. Therefore, unless one has a good reason for discounting a particular score as invalid, taking the highest of two or more presumably parallel test scores will lead to even more regression to the mean than would be observed by using just one score."
When he says "lead to more regression to the mean" at the end of this passage, he is referring to GT-identified students later testing below GT level.
So the author of the CogAT specifically does not recommend repeated administration, where any qualifying score gets and keeps you in, as a defensible criterion for GT entry. His recommendation for repeated testing is actually to average the scores, but use a lower cut score, and to make decisions on long-term GT placement only after consistently high averaged scores. Or to make GT placement more about state than trait. I.e., the presenting needs and skills of the student from year to year.
