Thursday, November 27, 2008

SAT Redux

I'm teaching an Introduction to Statistics course this term, and we just did linear regression. I used the opportunity to dig out a study of freshman grades, and showed the the students how to make a predictive model using SAT and high school GPA. The data was from several hundred students, with ACT mapped to SAT scores according to standard formulas. High school grades alone explained 33% of the variance in first year college grades in this sample. Adding SAT bumped it up to 37%. This isn't much added information, given all the trouble and expense of the SAT. Maybe it's just our type of students, but for us, it's not worth the effort. Better would be an attitude and behavior survey like the CIRP.

No comments:

Post a Comment