作者: Wen-Chung Wang , Po-Hsi Chen
关键词: Item analysis 、 Correlation 、 Data mining 、 Latent trait 、 Bayesian statistics 、 Item response theory 、 Rasch model 、 Computerized adaptive testing 、 Computer science 、 Bayesian probability
摘要: Multidimensional adaptive testing (MAT) procedures are proposed for the measurement of several latent traits by a single examination. Bayesian trait estimation and item selection derived. Simulations were conducted to compare efficiency MAT with those unidimensional random administration. The results showed that higher correlation between traits, more there were, scoring levels in items, efficient was than other two procedures. For tests containing multidimensional only is applicable, whereas not. Issues implementing discussed.