Diyorjon Abdullaev

Local item dependence (LID) refers to the situation where responses to items in a test or questionnaire are influenced by responses to other items in the test. This could be due to shared prompts, item content similarity, and deficiencies in item construction. LID due to a shared prompt is highly probable in cloze tests where items are nested within a passage. The purpose of this research is to examine the occurrence and magnitude of LID in a cloze test. A cloze test was analyzed with the Rasch model and locally dependent items were identified with the residual correlations. Findings showed that three pairs of items were locally dependent. When these items were removed from the analysis, test reliability dropped but item fit and unidimensionality improved. Removing the three locally dependent items did not affect person ability mean and standard deviation, though. The findings are discussed in terms of LID detection and modeling in the context of cloze test and language testing.

Cloze tests are commonly used in language testing as a quick measure of overall language ability or reading comprehension. A problem for the analysis of cloze tests with item response theory models is that cloze test items are locally dependent. This leads to the violation of the conditional or local independence assumption of IRT models. In this study, a new modeling strategy is suggested to circumvent the problem of local item dependence in cloze tests. This strategy involves identifying locally dependent items in the first step and combining them into polytomous items in the second step. Finally, partial credit model is applied to the combination of dichotomous and polytomous items. Our findings showed that the new strategy results in a better model-data fit than the dichotomous model where dependence is ignored but with a lower reliability. Results also indicated that the person and item parameters from the two models highly correlate. The findings are discussed in light of the literature on managing local dependence in educational tests.

Construct-irrelevant variance is considered as a major threat to validity which indicates the existence of additional unrelated variables that distort the meaning of test scores and cause the test to be biased. Differential item functioning (DIF) analysis is an important technique in examining the validity and fairness of educational tests. Concerning the importance of test fairness in large-scale exams, this study aimed to (1) detect gender DIF in the reading comprehension section of the B2 First exam using the Rasch model and Mantel-Haenszel method, and (2) investigate the comparability of results from the two DIF detection techniques. To this end, the reading section of the B2 First exam was administered to 207 undergraduate students of English as a foreign language (EFL). After checking the fit of the data to the Rasch model, the results of the Rasch model-based DIF analysis showed the presence of two items indicating DIF, whereas the results of Mantel-Haenszel showed that there were three gender-DIF items.