Educational and Psychological Measurement

This article compares several procedures in their efficacy for detecting differential item functioning (DIF): logistic regression analysis, the Mantel-Haenszel (MH) procedure, and the modified Mantel-Haenszel procedure by Mazor, Clauser, and Hambleton. It also compares the effect size measures that these procedures provide. In this study, different conditions of item parameters (difficulty and discrimination) and DIF magnitude were manipulated. Furthermore, both uniform and nonuniform DIF conditions were simulated. Results suggest that logistic regression analysis generally detected more items with DIF than the standard MH procedure and the modified MH procedure for symmetrical nonuniform DIF. The DIF effect size measures based on logistic regression, however, appeared to be insensitive to the specified DIF conditions.

The authors provide a cautionary note on reporting accurate eta-squared values from multifactor analysis of variance (ANOVA) designs. They reinforce the distinction between classical and partial eta-squared as measures of strength of association. They provide examples from articles published in premier psychology journals in which the authors erroneously reported partial eta-squared values as representing classical etasquared values. Finally, they discuss broader impacts of inaccurately reported etasquared values for theory development, meta-analytic reviews, and intervention programs.

Sample-size restrictions limit the contingency table approaches based on asymptotic distributions, such as the Mantel-Haenszel (MH) procedure, for detecting differential item functioning (DIF) in many practical applications. Within this framework, the present study investigated the power and Type I error performance of empirical and inferential criteria for DIF detection in small samples. Sample sizes (50/50, 100/50, 200/50, and 100/100 for the reference and focal groups, respectively), ability distributions (equal and unequal), and amount of DIF (moderate and high) were manipulated. The results show the advantages of employing theMHchi-square statistic using high levels of significance (• = .20) as opposed to the empirical criteria (cutoffs for categorizing DIF based on the magnitude of the MH common odds ratio estimator and the standardized p-difference statistic). Some considerations concerning Type I and Type II errors are made.

The Pearson correlation is used to depict effect sizes in the context of item response theory. Amultidimensional Rasch model is used to directly estimate the correlation between latent traits. Monte Carlo simulations were conducted to investigate whether the population correlation could be accurately estimated and whether the bootstrap method provided a good approximation of the distribution of the sample correlation. The independent variables were (a) the item response model, (b) test length, (c) sample size, and (d) the magnitude of correlation. The dependent variable was the empirical correlation coefficient. The results indicate that the population correlation can be accurately estimated and that the bootstrap method yields very a good approximation of the sampling distribution.

A total of 1,247 college students participated in this study on the effect of scale format on the reliability of Likert-type rating scales. The number of response categories ranged from 3 to 9. Anchor labels on the scales were provided for each response option or for the end points only. The results indicated that the scales with fewresponse categories tended to result in lower reliability, especially lower test-retest reliability. The scales with all the response options clearly labeled were likely to yield higher test-retest reliability than those with only the end points labeled. Scale design that leads to consistent participant responses as indicated by test-retest reliability should be preferred.

The Center for Epidemiologic Studies-Depression (CES-D) Scale is among the most commonly used measures of depressive symptomatology. Despite this, a paucity of research has been undertaken to examine the psychometric properties of responses to this scale. This meta-analytic study examined previously published studies of caregiving to identify factors that predict variance in reliability estimates (i.e., reliability generalization). The results suggest that the type of care recipient, the relationship to the care recipient, and CES-D Scale length each statistically affect reliability estimates. Only the number of items, however, appears to have a substantive effect. It is thus recommended that the original 20-item scale be used. Overall, it appears that responses to the CES-D Scale by care providers are largely reliable across these populations. The findings of an informal survey of authors suggest an incomplete awareness and appreciation for issues regarding reliability induction.

Recent lexical studies of personality structure suggest that there are six independent major dimensions of personality. TheHEXACOPersonality Inventory (HEXACO-PI), a new questionnaire that measures these six lexically derived personality constructs, was examined in two different cultural contexts using samples of 149 Francophone and 211 Korean respondents. Scores on the scales of the French and Korean versions of the HEXACO-PI were shown to have acceptable psychometric properties, including appropriate score distributions, high internal-consistency reliabilities, and low scale intercorrelations. In addition, the HEXACO-PI variables showed the expected pattern of correlations with markers of the Big Five and of lexical Honesty-Humility.

The purpose of the study was to investigate whether the six-factor structure of the Frost Multidimensional Perfectionism Scale could be replicated in a community-based sample. A sample of 255 adult participants (55.7% female, 44.3% male) ranging in age from 18 to 78 (mean = 37.0) completed the questionnaire. Based on the scree test and parallel analysis, four factors were selected for rotation. Varimax and oblimin rotation yielded four clearly interpretable subscales: Negative Projections, Organization, Parental Influences, and Achievement Expectations. Similarities between the original factor analysis and alternative solutions are discussed, as are the reasons for the suggested renaming of the factors.

This study examines the factor structure of scores on the English-language version of the Structure of Temperament Questionnaire. Scores from 300 college students were subjected to maximum-likelihood confirmatory factor analyses (CFA). A first-order model consisting of eight correlated factors and a second-order model consisting of two superordinate factors and eight first-order factors were tested. The results of the analyses indicate that the first-order correlated model is the best fit for the data. The English-language version of the Structure of Temperament Questionnaire can be used to measure eight correlated yet theoretically unique dimensions of temperament.

Three studies describe the development and refinement of a measure designed to assess Attitudes TowardWorking Single Parents (AWSP). Study 1 consisted of content validation of items written to assess respondent attitudes regarding the effect of single parenthood on two dimensions viewed as most central to the life experiences of single parents: work and family. Study 2 involved exploratory factor analysis and reliability analysis of the scores on the target measure. Finally, in Study 3, a confirmatory factor analysis was conducted to evaluate scale dimensionality, and discriminant, convergent, and subgroup validity coefficients were examined. The final scale may prove useful in guiding future research aimed at understanding the unique challenges faced byworking single parents.

The purpose of the study was to develop a short measure of internalized homophobia (IH), one that reflected contemporary attitudes toward homosexuality and included items designed to assess the domain of sexual comfort with gay men, a domain that has been notably absent from other measures of IH. The Short Internalized Homonegativity Scale (SIHS) was informed by the Reactions to Homosexuality Scale (RHS) and the contention that currently available measures of IH were outdated in their assessment of the construct and/or failed to assess its covert manifestations. A geographically diverse sample of gay men completed an online questionnaire (N = 1,307), and the 677 respondents from the United States formed the sample for the study. Confirmatory factor analyses supported a single higher order construct of IH comprising the lower order factors of Public Identification as Gay, Sexual Comfort With Gay Men, and Social Comfort With Gay Men.