COMPARING EQUATING ERRORS ON VARIOUS FACTORS FOR SUBTESTS WHICH HAVE ADDED VALUE.

Uçar, Arzu; Sünbül, Önder

In this study, 270 different data sets were generated through R 3.1.1. program as dichotomous data for X and Y test forms, each consisting of two sub-tests, according to the two-parameter logistic model (2PLM). Forms include an anchor test, which consists of two sub-tests like test form. The sub-test length of the anchor test is 40% of the total test. For sample sizes of 20, 25, 50, 100, 200, 500, X and Y test forms were created with correlation levels between sub-tests of 0.70, 0.80, 0.90; average difficulty level differences between test forms of 0.0, 0.40, 0.70; sub-test lengths of 10, 15, 30, 50, 80. Sub-tests were equated as a result of 100 replications using identity, linear-chain, Braun/Holland, and circular-arc equating methods. The results obtained from this simulation study were evaluated based on equating error. The findings indicated that in the case when sample size was 100 and more, subtest length was 10, 15 and 30 and the level of average difficulty difference between form 0.0, it was concluded that equating forms would give better results than not equating. Furthermore, the circle-arc method was found to have less equating error than other methods under most of the conditions studied.

LOGISTIC regression analysis; ANCHOR testing; SAMPLE size (Statistics); STATISTICAL sampling; SIMULATION methods & models

Journal of Advanced Education Studies, 2024, Vol 6, Issue 1, p92

2687-3087

Academic Journal

10.48166/ejaes.1438652