Comparison of Different Forms of a Test with or without Items that Exhibit DIF
MetadataShow full item record
Purpose: This research investigates Tests that include DIF items and which are purified from DIF items. While doing this, the ability estimations and purified DIF items are compared to understand whether there is a correlation between the estimations. Method: The researcher used to R 3.4.1 in order to compare the items and after this situation; according to manipulated factors, we carried out the data production under different circumstances with the help of simulation study. The manipulated factors were determined levels of sample size (1000, 2000), test length (40, 60) and percentage of DIF (%5, %10). By using the new data each condition’s DIF items’ ability estimations were carried out. Afterward, DIF items purified from the tests and later the abilities were estimated. The correlation between the ability parameters was calculated by using the Spearman's Rank Correlation Coefficient and these parameters were calculated separately according to the eight conditions. Findings: After calculations, all of the coefficients of correlations (rs)’ values were almost zero (p<0.01). In other words the test length 40 and 60, sample size 1000 and 2000, percentage of DIF %5 and %10, when we crossed these parameters in different eight conditions, there was no familiar correlation between the tests that include DIF items and tests of that purified from DIF items. Besides, there was no correlation between the tests thinking the ability estimations; if we exclude DIF items from the tests, the individuals’ test ranking changes, too. Implication for Research and Practice: This study showed that tests that include DIF items affect the ability estimation of individuals. In the frame of this result, teachers, administrators, and policymakers should bear in mind tests DIF potential. Also, this study may be carried out by using various conditions.
SourceEurasian Journal of Educational Research