WISC-III簡式「多重版本－多重樣本－多重方法」之智商估計效度研究

The Wechsler Intelligence Scale, Third Edition (WISC-III) is one of the most important individualized IQ test in Taiwan. However, its' long testing time has been a bothersome factor. Study with「Multi-Form/ Multi-Sample/Multi-Method」 design was conducted to investigate the clinical validities for possible short form solutions. A total of 5 short forms, 5 groups of independent samples, and 3 different statistic methods were combined. For each combination, the correlation between estimated IQ and the full scaled IQ (FSIQ); the base rate of the difference between the estimated IQ and the FSIQ; and the accuracy for classification were all calculated. Major findings were: (1) linear regression is the best among all three statistic methods, which shows the best accuracy in FSIQ estimation; (2) The correlations between estimated IQ and the FSIQ are high for almost all combinations (majority are in the .90s), however, large difference does exist between those two scores. Decisions for short form selection thus could not be made based on the correlation coefficient only. (3) The performance of each short forms varies among different children samples, in order to ensure the accuracy for FSIQ estimation, it was suggested that different short form could be used for different clinical sample in the future. However, at this moment,「Vocabulary、Similarity 、Block Design、Picture Arrangement」 is the best choice for groups with high heterogeneity. The combination of「 Similarity、Arithmetic、Block Deign、Picture Completion」 could also be considered while the scoring convenience become the top consideration. Nonetheless, practitioners have to realize that there are no perfect short forms, error does exist whenever a short form was used to estimate FSIQ. Thus, in order not to hurt the child being tested, short form should be used and explained in an extremely careful manner.