Abstract:
The purposes of the study were to compare test results analyzed by two different models, namely the Classical and the Rasch; with respect to number of items selected, the reliability, the validity, and the correlation of the total scores obtained from the selected items of the two models. There are 1884 Mathayom Suksa One students taking two tests of entrance examination, the Analogy Test and the Science Test, at the Bangkapi and Satree Settabutbumpen Schools. These tests are the standardized tests produced by the Educational and Psychological Testing Bureau of Sri Nakarintarawirote Prasanmit University. The Analogy test scores were used for the main analysis and the Science test scores were the criterion used for the con-current validity purpose. The CIA (Classical Item Analysis) and BICAL Computer Program were used for item analysis of the two models. The KR#20 reliability, the concurrent validity and the correlation of total scores resulting from two types of models were shown by the Pearson Product Moment and transformed by the fishers Z and tested by Z Test. The findings were as follows: (1) The two models produced significant different in numbers of test items. (p < .05) (2) The Classical Model produced higher value of reliability than that of the Rash Model. (p < .01) (3) There was no significant difference between the con-current validity obtained by the two models. (4) There was a significant correlation between total scores obtained by selected items of the two models. (p < .001)