Comments on Holland and Wainer's "Sources of Uncertainty Often Ignored in Adjusting State Mean SAT scores for Differential Participation Rates: The Rules of the Game"
Don Edwards, Cynthia Cummings · Applied Measurement in Education · 1990
Probability selection models for Scholastic Achievement Test (SAT) test takers using truncated normal distributions have been independently discussed by Taube and Linden (1989) and Edwards and Beckworth (1989). Holland and Wainer (1990) provided one formalization of the truncated normal selection model, and rejected it based on patterns in state test-taker variances and test score histograms. Their model, however, is fundamentally different from that considered by Edwards and Beckworth; the former hypothesizes that the probability of taking the test is a function of test score, the latter that it is a function of true score. Here, an evolved form of the Edwards and Beckworth model is outlined, and its relationship with earlier models is discussed. It is shown that the arguments of Holland and Wainer are not sufficient to reject this model.