| Abstract: | Developed a computer simulation procedure to reproduce the overall pattern of results of the Educational Testing Service 1971 growth study. Then, simulated data for 7 sets of 10,000 to 15,000 cases were analyzed with several techniques for assessing growth. Techniques were compared on the basis of correlations between estimated and true growth scores and of root mean square errors. Growth was estimated most accurately by procedures that involved the pretest-posttest difference, and for practical purposes all estimates that involved this difference had approximately equal accuracy. In particular, the simple difference between pre- and posttest scores seemed about as accurate as any other estimate, was easier to compute, and should be meaningful to nonresearchers. (21 ref) (PsycINFO Database Record (c) 2010 APA, all rights reserved) |
