Detecting Differential Item Functioning Using Logistic Regression Procedures

A logistic regression model for characterizing differential item functioning (DIF) between two groups is presented. A distinction is drawn between uniform and nonuniform DIF in terms of the parameters of the model. A statistic for testing the hypothesis of no DIF is developed. Through simulation studies, it is shown that the logistic regression procedure is more powerful than the Mantel-Haenszel procedure for detecting nonuniform DIF and as powerful in detecting uniform DIF.     

Logistic Regression Procedure Using Penalized Maximum Likelihood Estimation for Differential Item Functioning

In the logistic regression (LR) procedure for differential item functioning (DIF), the parameters of LR have often been estimated using maximum likelihood (ML) estimation. However, ML estimation suffers from the finite-sample bias. Furthermore, ML estimation for LR can be substantially biased in the presence of rare event data. The bias of ML estimation due to small samples and rare event data can degrade the performance of the LR procedure, especially when testing the DIF of difficult items in small samples. Penalized ML (PML) estimation was originally developed to reduce the finite-sample bias of conventional ML estimation and also was known to reduce the bias in the estimation of LR for the rare events data. The goal of this study is to compare the performances of the LR procedures based on the ML and PML estimation in terms of the statistical power and Type I error. In a simulation study, Swaminathan and Rogers's Wald test based on PML estimation (PSR) showed the highest statistical power in most of the simulation conditions, and LRT based on conventional PML estimation (PLRT) showed the most robust and stable Type I error. The discussion about the trade-off between bias and variance is presented in the discussion section.     

Logistic Regression and Its Use in Detecting Differential Item Functioning in Polytomous Items

A computer simulation study was conducted to determine the feasibility of using logistic regression procedures to detect differential item functioning (DIF) in polytomous items. One item in a simulated test of 25 items contained DIF; parameters' for that item were varied to create three conditions of nonuniform DIF and one of uniform DIF. Item scores were generated using a generalized partial credit model, and the data were recoded into multiple dichotomies in order to use logistic regression procedures. Results indicate that logistic regression is powerful in detecting most forms of DIF; however, it required large amounts of data manipulation, and interpretation of the results was sometimes difficult. Some logistic regression procedures may be useful in the post hoc analysis of DlF for polytomous items.     

Using Logistic Regression and the Mantel-Haenszel With Multiple Ability Estimates to Detect Differential Item Functioning

Logistic regression has recently been advanced as a viable procedure for detecting differential item functioning (DIF). One of the advantages of this procedure is the considerable flexibility it offers in the specification of the regression equation. This article describes incorporating two ability estimates into a single regression analysis, with the result that substantially fewer items exhibit DIF. A comparable analysis is conducted using the Mantel-Haenszel with similar results. It is argued that by simultaneously conditioning on two relevant ability estimates, more accurate matching of examinees in the reference and focal groups is obtained, and thus multidimensional item impact is not mistakenly identified as DIF.     

Effect Size Measures for Differential Item Functioning in a Multidimensional IRT Model

This study adapted an effect size measure used for studying differential item functioning (DIF) in unidimensional tests and extended the measure to multidimensional tests. Two effect size measures were considered in a multidimensional item response theory model: signed weighted P‐difference and unsigned weighted P‐difference. The performance of the effect size measures was investigated under various simulation conditions including different sample sizes and DIF magnitudes. As another way of studying DIF, the χ² difference test was included to compare the result of statistical significance (statistical tests) with that of practical significance (effect size measures). The adequacy of existing effect size criteria used in unidimensional tests was also evaluated. Both effect size measures worked well in estimating true effect sizes, identifying DIF types, and classifying effect size categories. Finally, a real data analysis was conducted to support the simulation results.     

Simulated Tests of Differential Item Functioning Using SIBTEST With and Without Impact

Monte Carlo simulations with 20,000 replications are reported to estimate the probability of rejecting the null hypothesis regarding DIF using SIBTEST when there is DIF present and/or when impact is present due to differences on the primary dimension to be measured. Sample sizes are varied from 250 to 2000 and test lengths from 10 to 40 items. Results generally support previous findings for Type I error rates and power. Impact is inversely related to test length. The combination of DIF and impact, with the focal group having lower ability on both the primary and secondary dimensions, results in impact partially masking DIF so that items biased toward the reference group are less likely to be detected.     

Linguistic Complexity,Schematic Representations,and Differential Item Functioning for English Language Learners in Math Tests

This article examines nonmathematical linguistic complexity as a source of differential item functioning (DIF) in math word problems for English language learners (ELLs). Specifically, this study investigates the relationship between item measures of linguistic complexity, nonlinguistic forms of representation and DIF measures based on item response theory difficulty parameters in a state fourth-grade math test. This study revealed that the greater the item nonmathematical lexical and syntactic complexity, the greater are the differences in difficulty parameter estimates favoring non-ELLs over ELLs. However, the impact of linguistic complexity on DIF is attenuated when items provide nonlinguistic schematic representations that help ELLs make meaning of the text, suggesting that their inclusion could help mitigate the negative effect of increased linguistic complexity in math word problems.     

Differential Item Functioning for Minority Examinees on the SAT

The standardization approach to assessing differential item functioning (DIF), including standardized distractor analysis, is described. The results of studies conducted on Asian Americans, Hispanics (Mexican Americans and Puerto Ricans), and Blacks on the Scholastic Aptitude Test (SAT) are described and then synthesized across studies. Where the groups were limited to include only examinees who spoke English as their best language, very few items across forms and ethnic groups exhibited large DIF. Major findings include evidence of differential speededness (where minority examinees did not complete SAT-Verbal sections at the same rate as White students with comparable SAT-Verbal scores) for Blacks and Hispanics and, when the item content is of special interest, advantages for the relevant ethnic group. In addition, homographs tend to disadvantage all three ethnic groups, but the effect of vertical relationships in analogy items are not as consistent. Although these findings are important in understanding DIF, they do not seem to account for all differences. Other variables related to DIF still need to be identified. Furthermore, these findings are seen as tentative until corroborated by studies using controlled data collection designs.     

Differential Item Functioning Analysis for Accommodated Versus Nonaccommodated Students

The No Child Left Behind act resulted in an increased reliance on large-scale standardized tests to assess the progress of individual students as well as schools. In addition, emphasis was placed on including all students in the testing programs as well as those with disabilities. As a result, the role of testing accommodations has become more central in discussions about test fairness and accessibility as well as evidence of validity. This study seeks to examine whether there exists differential item functioning for math and language items between special education examinees receiving accommodations and those not receiving accommodations.     

Identifying Sources of Differential Item and Bundle Functioning on Translated Achievement Tests: A Confirmatory Analysis

Increasingly, tests are being translated and adapted into different languages. Differential item functioning (DIF) analyses are often used to identify non-equivalent items across language groups. However, few studies have focused on understanding why some translated items produce DIF. The purpose of the current study is to identify sources of differential item and bundle functioning on translated achievement tests using substantive and statistical analyses. A substantive analysis of existing DIF items was conducted by an 11-member committee of testing specialists. In their review, four sources of translation DIF were identified. Two certified translators used these four sources to categorize a new set of DIF items from Grade 6 and 9 Mathematics and Social Studies Achievement Tests. Each item was associated with a specific source of translation DIF and each item was anticipated to favor a specific group of examinees. Then, a statistical analysis was conducted on the items in each category using SIBTEST. The translators sorted the mathematics DIF items into three sources, and they correctly predicted the group that would be favored for seven of the eight items or bundles of items across two grade levels. The translators sorted the social studies DIF items into four sources, and they correctly predicted the group that would be favored for eight of the 13 items or bundles of items across two grade levels. The majority of items in mathematics and social studies were associated with differences in the words, expressions, or sentence structure of items that are not inherent to the language and/or culture. By combining substantive and statistical DIF analyses, researchers can study the sources of DIF and create a body of confirmed DIF hypotheses that may be used to develop guidelines and test construction principles for reducing DIF on translated tests.     

Investigating Linguistic Sources of Differential Item Functioning Using Expert Think-Aloud Protocols in Science Achievement Tests

Even if national and international assessments are designed to be comparable, subsequent psychometric analyses often reveal differential item functioning (DIF). Central to achieving comparability is to examine the presence of DIF, and if DIF is found, to investigate its sources to ensure differentially functioning items that do not lead to bias. In this study, sources of DIF were examined using think-aloud protocols. The think-aloud protocols of expert reviewers were conducted for comparing the English and French versions of 40 items previously identified as DIF (N?=?20) and non-DIF (N?=?20). Three highly trained and experienced experts in verifying and accepting/rejecting multi-lingual versions of curriculum and testing materials for government purposes participated in this study. Although there is a considerable amount of agreement in the identification of differentially functioning items, experts do not consistently identify and distinguish DIF and non-DIF items. Our analyses of the think-aloud protocols identified particular linguistic, general pedagogical, content-related, and cognitive factors related to sources of DIF. Implications are provided for the process of arriving at the identification of DIF, prior to the actual administration of tests at national and international levels.     

Detection of Differential Item Functioning with Nonlinear Regression: A Non‐IRT Approach Accounting for Guessing

In this article we present a general approach not relying on item response theory models (non‐IRT) to detect differential item functioning (DIF) in dichotomous items with presence of guessing. The proposed nonlinear regression (NLR) procedure for DIF detection is an extension of method based on logistic regression. As a non‐IRT approach, NLR can be seen as a proxy of detection based on the three‐parameter IRT model which is a standard tool in the study field. Hence, NLR fills a logical gap in DIF detection methodology and as such is important for educational purposes. Moreover, the advantages of the NLR procedure as well as comparison to other commonly used methods are demonstrated in a simulation study. A real data analysis is offered to demonstrate practical use of the method.     

两种DIF检测方法的模拟研究

本研究通过Monte Carlo模拟,探讨MH和LR两种方法在检测DIF时I型错误率和检出率的情况。实验结果表明两种方法的I型错误均控制在0.05左右(α=0.05),LR方法的I型错误率呈现出更加稳定的状态。一致性DIF时,MH方法的检出率略高于LR方法;而非一致性DIF时,LR方法的检出率大大高于MH方法,MH方法对非一致性DIF不敏感。另外,两种方法一致性DIF的检出率随有DIF题目的比例增加而增加,而非一致性DIF的检出率随比例的增加而有所降低。     

An Analytic Comparison of Effect Sizes for Differential Item Functioning

Three types of effects sizes for DIF are described in this exposition: log of the odds-ratio (differences in log-odds), differences in probability-correct, and proportion of variance accounted for. Using these indices involves conceptualizing the degree of DIF in different ways. This integrative review discusses how these measures are impacted in different ways by item difficulty, item discrimination, and item lower asymptote. For example, for a fixed discrimination, the difference in probabilities decreases as the difference between the item difficulty and the mean ability increases. Under the same conditions, the log of the odds-ratio remains constant if the lower asymptote is zero. A non-zero lower asymptote decreases the absolute value of the probability difference symmetrically for easy and hard items, but it decreases the absolute value of the log-odds difference much more for difficult items. Thus, one cannot set a criterion for defining a large effect size in one metric and find a corresponding criterion in another metric that is equivalent across all items or ability distributions. In choosing an effect size, these differences must be understood and considered.     

Sample Size Calculations for Precise Interval Estimation of the Eta-Squared Effect Size

Analysis of variance is one of the most frequently used statistical analyses in the behavioral, educational, and social sciences, and special attention has been paid to the selection and use of an appropriate effect size measure of association in analysis of variance. This article presents the sample size procedures for precise interval estimation of eta-squared and partial eta-squared in fixed-effects analysis of variance designs. The desired precision of a confidence interval is assessed with respect to (a) the control of expected width and (b) the tolerance probability of interval width within a designated value. In addition, sample size calculations for standardized contrasts of treatment effects and corresponding partial strength of association effect sizes are also considered.     

Detection of Differential Item Functioning in Multiple Groups

Detection of differential item functioning (DIF) is most often done between two groups of examinees under item response theory. It is sometimes important, however, to determine whether DIF is present in more than two groups. In this article we present a method for detection of DIF in multiple groups. The method is closely related to Lard's chi-square for comparing vectors of item parameters estimated in two groups. An example using real data is provided.     

A Mixture Model Analysis of Differential Item Functioning

Once a differential item functioning (DIF) item has been identified, little is known about the examinees for whom the item functions differentially. This is because DIF focuses on manifest group characteristics that are associated with it, but do not explain why examinees respond differentially to items. We first analyze item response patterns for gender DIF and then illustrate, through the use of a mixture item response theory (IRT) model, how the manifest characteristic associated with DIF often has a very weak relationship with the latent groups actually being advantaged or disadvantaged by the item(s). Next, we propose an alternative approach to DIF assessment that first uses an exploratory mixture model analysis to define the primary dimension(s) that contribute to DIF, and secondly studies examinee characteristics associated with those dimensions in order to understand the cause(s) of DIF. Comparison of academic characteristics of these examinees across classes reveals some clear differences in manifest characteristics between groups.     

Differential Item Functioning on the SAT-M Braille Edition

This study attempted to pinpoint the causes of differential item difficulty for blind students taking the braille edition of the Scholastic Aptitude Test's Mathematical section (SAT-M). The study method involved reviewing the literature to identify factors that might cause differential item functioning for these examinees, forming item categories based on these factors, identifying categories that functioned differentially, and assessing the functioning o f the items comprising deviant categories to determine if the differential effect was pervasive. Results showed an association between selected item categories and differential functioning, particularly for items that included figures in the stimulus, items for which spatial estimation was helpful in eliminating at least two of the options, and items that presented figures that were small or medium in size. The precise meaning of this association was unclear, however, because some items from the suspected categories functioned normally, factors other than the hypothesized ones might have caused the observed aberrant item behavior, and the differential difficulty might reflect real population differences in relevant content knowledge     

Differential Item Functioning Effect Size From the Multigroup Confirmatory Factor Analysis for a Meta-Analysis: A Simulation Study

This study presents a new approach to synthesizing differential item functioning (DIF) effect size: First, using correlation matrices from each study, we perform a multigroup confirmatory factor analysis (MGCFA) that examines measurement invariance of a test item between two subgroups (i.e., focal and reference groups). Then we synthesize, across the studies, the differences in the estimated factor loadings between the two subgroups, resulting in a meta-analytic summary of the MGCFA effect sizes (MGCFA-ES). The performance of this new approach was examined using a Monte Carlo simulation, where we created 108 conditions by four factors: (1) three levels of item difficulty, (2) four magnitudes of DIF, (3) three levels of sample size, and (4) three types of correlation matrix (tetrachoric, adjusted Pearson, and Pearson). Results indicate that when MGCFA is fitted to tetrachoric correlation matrices, the meta-analytic summary of the MGCFA-ES performed best in terms of bias and mean square error values, 95% confidence interval coverages, empirical standard errors, Type I error rates, and statistical power; and reasonably well with adjusted Pearson correlation matrices. In addition, when tetrachoric correlation matrices are used, a meta-analytic summary of the MGCFA-ES performed well, particularly, under the condition that a high difficulty item with a large DIF was administered to a large sample size. Our result offers an option for synthesizing the magnitude of DIF on a flagged item across studies in practice.