• Version
• Download 128
• File Size 357.66 KB
• File Count 1
• Create Date August 2, 2018
• Last Updated August 2, 2018

STUDY OF FAIRNESS IN THE PHILIPPINE APTITUDE CLASSIFICATION TEST

Test fairness, according to Standards, has four characteristics. These are equitable treatment of all examinees in the testing process, the absence of bias, the equality of testing outcomes for examinee subgroups and equity in opportunity to learn the material covered in an achievement test.The development of the Philippine Aptitude Classification Test (PACT) considered giving fair treatment to all examinees in terms of context, purpose of testing and the manner in which the test scores were used. The absence of bias, however, was not examined. Hence, the purpose of this study is to gather baseline data on the extent to which PACT is an unbiased instrument using the three widely used differential item functioning (DIF) procedures: Rasch Model (RM), Mantel-Haenszel procedure (MH) and the Logistic Regression procedure (LR). A sample of 2,296 examinees was drawn from a total of 52,006 high school examinees who took the PACT in SY 2007-08. The sample had an equal ratio of male and female examinees while the ratio of Metro Manila and non-Metro Manila group was approximately, six is to four.At the test level, the study showed that the differences in scores between gender groups and between geographic groups are very minimal with scores ranging from 10% to 16% of one (1) standard deviation. On the item level, a large number of items, with respect to gender (73 out of 240 items) and geographic groups (47 out of 240 items) displayed DIF for the b-parameter (difficulty) (p <, .01). The presence of DIF on these items, however, is not sufficient evidence to conclude that the items are biased. The results of the study recommends further substantive investigation, like a replication, to determine if performance differences that have been observed are, in fact, due to impact and not due to bias in the items.

Attached Files