I think it is totally idiotic for the school to use the CogAT to separate out levels within a gifted population, unless the school is administering the test *way* out of level. The CogAT is merely a screening test, though most district officials who make decisions based on it don't seem to know the first thing about it.
Personally, if it were up to me, I'd "stratify" by observed achievement.
I completely agree with Kai. The CogAT is a GT screener for potential or likelihood of GT. Clustering, I would guess, is most successful based on achievement.
