Differential Item Functioning: Methods and applications (27)

Chair: Brian Clauser, Wednesday 22nd July, 15.25 - 16.45, Lowercroft, School of Pythagoras. 

Marie Wiberg, Department of Statistics, Umeå University, Sweden. Detecting Differential Item Functioning in licensure tests. (182)

Ya-Hui Su, Berkeley Evaluation and Assessment Research (BEAR) Center, University of California, Berkeley,USA and Wen-Chung Wang, Department of Educational Psychology, Counselling and Learning Needs, The Hong Kong Institute of Education. The ‘DIF-free-then-DIF’ strategy applied to the logistic regression procedure for DIF assessment. (243)

Ahyoung Kim and Ji Eun Joo, Ewha Womans University, Seoul, Korea. Group Differences in Strategy Use in Solving Spatial Tasks (254)

ABSTRACTS 

Detecting Differential Item Functioning in licensure tests. (182)
Marie Wiberg
To obtain a license it is common to use a test with a cutoff score where test takers with scores at or above the cutoff score obtain a license and test takers below the cut‐off score does not obtain a license. Since it is crucial how items behave for test takers with results close to the cut off score it is of interest to pay special attention to DIF in this area. One way to examine DIF is to classify the test score and examine the DIF with common methods in each interval. Another promising method is to model the items with log linear models where one immediately uses DIF in intervals. This study gives reasons behind such approach as well as compares it with traditional DIF methods including when the test scores are classified into intervals.

The ‘DIF-free-then-DIF’ strategy applied to the logistic regression procedure for DIF assessment. (243)
Ya-Hui Su and Wen-Chung Wang
In this study, we apply the DIF-free-then DIF strategy (Wang, 2008) to the logistic regression method for assessment of differential item functioning (DIF). The strategy involves two steps. In the 1st step, a set of items that are the least likely to have DIF are selected by using the iterative logistic regression method for DIF-free items (LR-f). In the 2nd step, these selected items serve as the matching variable and the other items are assessed for DIF by using the logistic regression method with constant items (LR-c; Su & Wang, 2009). A simulation study was conducted to assess the performance of the strategy. The results show that the strategy outperformed the traditional logistic regression method and the logistic regression method with scale purification, when tests contained a high percentage of DIF items. An empirical example is given.

Group Differences in Strategy Use in Solving Spatial Tasks (254)
Ahyoung Kim and Ji Eun Joo
It is widely recognized that individual differences exist in the use of solution strategies, and that different strategies for different types of tasks are used when solving spatial tasks. The present study aimed to identify the characteristics of individuals that adopt different strategies, as well as the types of item formats that require the use of different strategies. A spatial ability test consisting of eight types of item formats, along with five alternative solution strategies attached to each item, was administered to 1,063 university students. The items were assumed to represent different spatial factors, and the solution strategies included holistic, analytic, and three others. To investigate group differences, participants were classified by latent class, sex, and spatial ability levels. Two latent classes for each item type have been identified in latent class analyses. The results revealed that the analytic strategy was used predominantly in picture completion and embedded pieces items in both latent class groups and in high spatial ability groups, whereas the holistic strategy was predominantly used in 2- and 3-D rotation, form development, and paper folding items. Similar results were obtained in the comparison between high and low ability groups. No differences in sex were found.