Comparison of Johnson-Neyman, Mantel-Haenszel and partial correlation procedures for detecting differential item performance for moderate sample size /
| Main Author: | |
|---|---|
| Other Authors: | , , |
| Format: | Thesis Book |
| Language: | English |
| Published: |
1991.
|
| Subjects: | |
| Online Access: | Link to OAKTrust copy |
| Abstract: | Three statistical techniques for detecting differential item performance (DIP) namely, the partial correlation (PC), the Mantel-Haenszel (MH) method, and the Johnson-Neyman (JN) procedure, were empirically compared for moderate sample size. In a Monte Carlo study the three techniques were compared for cases based on two different number of items (25, 50), two sample sizes (50, 100), three types of bias (uniform bias, intercept; non-uniform bias, slope; and both bias combined, intercept-slope), and two levels of bias (moderate, high). A total of 28 test conditions were each replicated 1000 times from simulation data generated by computer. Actual data from a standardization sample were analyzed. From the results it appears that the JN technique is more conservative under the null hypothesis case, while being more powerful than the PC and the MH procedures in detecting non-uniform (slope) bias. However, in case of uniform (intercept) bias, the JN seems to be as powerful as the PC and the MH procedures. When intercept and slope bias are combined, use of both the JN and the MH procedures may yield better results. Further, the JN technique is more informative than the other two methods in determining regions of significance across the ability dimension. |
|---|---|
| Item Description: | Typescript (photocopy). Vita. "Major subject: Educational Psychology." |
| Physical Description: | xi, 132 leaves : illustrations ; 29 cm |
| Bibliography: | Includes bibliographical references. |