Key points are not available for this paper at this time.
This paper examines several structural models of similarity between a battery of conventional tests and a battery of computerized adaptive tests designed to measure the same aptitudes. Twelve plausible models are reviewed and are fitted to sample data in a double cross-validation design. Three of the 12 models provided reasonable summaries of the data. One model with a multiplicative structure (Browne, 1984) performed quite well. This model provides an estimate of the disattenuated method correlation between conventional testing and adaptive testing. In the current data, this correlation was estimated to be 0.97 and 0.98 in the two halves of the data. These results were interpreted as very supportive of the CAT tests as replacements for the conventional tests.
Robert Cudeck (Mon,) studied this question.