David J. Hessen

Stereotype Threat and Group Differences in Test Performance: A Question of Measurement Invariance

Studies into the effects of stereotype threat (ST) on test performance have shed new light on race and sex differences in achievement and intelligence test scores. In this article, the authors relate ST theory to the psychometric concept of measurement invariance and show that ST effects may be viewed as a source of measurement bias. As such, ST effects are detectable by means of multi-group confirmatory factor analysis. This enables research into the generalizability of ST effects to real-life or high-stakes testing. The modeling approach is described in detail and applied to 3 experiments in which the amount of ST for minorities and women was manipulated. Results indicate that ST results in measurement bias of intelligence and mathematics tests.

Age and gender differences in depression across adolescence: real or ‘bias’?

BACKGROUND Since developmental psychologists are interested in explaining age and gender differences in depression across adolescence, it is important to investigate to what extent these observed differences can be attributed to measurement bias. Measurement bias may arise when the phenomenology of depression varies with age or gender, i.e., when younger versus older adolescents or girls versus boys differ in the way depression is experienced or expressed. METHODS The Childrens Depression Inventory (CDI) was administered to a large school population (N = 4048) aged 8-17 years. A 4-factor model was selected by means of factor analyses for ordered categorical measures. For each of the four factor scales measurement invariance with respect to gender and age (late childhood, early and middle adolescence) was tested using item response theory analyses. Subsequently, to examine which items contributed to measurement bias, all items were studied for differential item functioning (DIF). Finally, it was investigated how developmental patterns changed if measurement biases were accounted for. RESULTS For each of the factors Self-Deprecation, Dysphoria, School Problems, and Social Problems measurement bias with respect to both gender and age was found and many items showed DIF. Developmental patterns changed profoundly when measurement bias was taken into account. The CDI seemed to particularly overestimate depression in late childhood, and underestimate depression in middle adolescent boys. CONCLUSIONS For scientific as well as clinical use of the CDI, measurement bias with respect to gender and age should be accounted for.

Heteroscedastic one-factor models and marginal maximum likelihood estimation

In the present paper, a general class of heteroscedastic one-factor models is considered. In these models, the residual variances of the observed scores are explicitly modelled as parametric functions of the one-dimensional factor score. A marginal maximum likelihood procedure for parameter estimation is proposed under both the assumption of multivariate normality of the observed scores conditional on the single common factor score and the assumption of normality of the common factor score. A likelihood ratio test is derived, which can be used to test the usual homoscedastic one-factor model against one of the proposed heteroscedastic models. Simulation studies are carried out to investigate the robustness and the power of this likelihood ratio test. Results show that the asymptotic properties of the test statistic hold under both small test length conditions and small sample size conditions. Results also show under what conditions the power to detect different heteroscedasticity parameter values is either small, medium, or large. Finally, for illustrative purposes, the marginal maximum likelihood estimation procedure and the likelihood ratio test are applied to real data.

The multigroup common factor model with minimal uniqueness constraints and the power to detect uniform bias

An alternative formulation of the multigroup common factor model with minimal uniqueness constraints is considered. This alternative formulation is based on a simple identification constraint that is related to the standard maximum likelihood constraint used in single-group common factor analysis. It is argued that the alternative formulation leads to less technical difficulties in applications than earlier formulations of this multigroup common factor model. Furthermore, associated tests for various measurement invariance constraints across groups are proposed, such as an omnibus test for the absence of uniform bias. By means of an empirical example, the fitting of several multigroup common factor models with minimal uniqueness constraints and the testing for measurement invariance over groups are demonstrated. The nesting of multigroup confirmatory factor models under the multigroup common factor model with minimal uniqueness constraints is also discussed. Finally, a small study is performed to investigate the drop in power to detect uniform bias in using the multigroup common factor model with minimal uniqueness constrains instead of a confirmatory special case. The results of the study show a small drop in power under all research conditions.

Body Satisfaction and Sexual Health in Dutch Female University Students

Studies in the field of body image have primarily addressed its negative aspects, such as body dissatisfaction. The present study focused instead on women who are satisfied with their bodies and on how body satisfaction relates to sexual health. A sample of 319 Dutch female university students completed an online survey that included items about body image evaluation, body image investment, overweight preoccupation, body image affect during sexual activity, sexual frequency, sexual functioning, and sexual self-esteem. We found that the level of body dissatisfaction was minimal in our sample. The majority reported neutral or mildly positive body evaluations, and in 30% of the sample these evaluations were clearly positive. Comparisons between women who reported positive versus neutral body evaluations showed that the body-satisfied women had lower body mass indexes (BMIs) and reported less body image investment, less overweight preoccupation, and less body self-consciousness during sexual activity. With regard to sexual health, they reported higher sexual self-esteem and better sexual functioning. Furthermore, we found that body image self-consciousness was negatively associated with sexual functioning, sexual self-esteem, and frequency of sexual activity with a partner. Body satisfaction did not account for a portion of the relationship of body self-consciousness during sexual activity with sexual health.

Clinical Usefulness of the Kiddie-Disruptive Behavior Disorder Schedule in the Diagnosis of DBD and ADHD in Preschool Children

The aim of the present study was to investigate the clinical usefulness of a semi-structured diagnostic parent interview, i.e., the Kiddie-Disruptive Behavior Disorder Schedule (K-DBDS), in preschool children. For Oppositional Defiant Disorder (ODD), to define symptoms two coding methods were compared, i.e., one based on the threshold “often” and the other based on the frequency of behaviors in combination with the presence of clinical concern. For Attention-Deficit/Hyperactivity Disorder (ADHD), to define symptoms, two coding methods were compared, i.e., one with and one without consideration of pervasiveness across contexts. Participants were referred preschool children with externalizing behavioral problems (N = 193; 83% male) and typically developing (TD) children (N = 58; 71% male). The referred children were given a diagnosis of either ODD/CD (N = 39), or ADHD (N = 58) or comorbid ODD/CD+ADHD (N = 57) or no diagnosis (N = 39) based on best-estimate diagnosis. Receiver Operating Characteristic curve analyses showed that a cutoff score of four ODD symptoms using “often” as the threshold for frequency of behaviors led to a sensitivity of 87% and a specificity of 93%; the coding method which included the frequency of behaviors yielded a sensitivity of 56% and a specificity of 100%. For ADHD, a clinical cutoff score of five symptoms without the pervasiveness criterion yielded a sensitivity of 83% and a specificity of 98%; when the pervasiveness criterion was included sensitivity was 77% and specificity 98%. In the clinical assessment of ODD and ADHD in preschool children, the K-DBDS may be used with ODD symptom definition based on the threshold “often” and ADHD pervasiveness across contexts not included.

Clinical Usefulness of Observational Assessment in the Diagnosis of DBD and ADHD in Preschoolers

The aim of the present study was to investigate the clinical usefulness of an observational tool—the Disruptive Behavior Diagnostic Observation Schedule (DB-DOS)—in the diagnosis of disruptive behavior disorders (DBD) and attention deficit/hyperactivity disorder (ADHD) in preschoolers. We hypothesized that the DB-DOS may help support the presumption of a diagnosis generated by the information from parents and teachers (or other caregivers). Participants were referred preschool children with externalizing behavioral problems (N = 193; 83% male) and typically developing children (N = 58; 71% male). In view of the clinical validity study each child was given a diagnosis of either DBD (N = 40), or ADHD (N = 54) or comorbid (DBD + ADHD; N = 66) based on best-estimate diagnosis. The DB-DOS demonstrated good interrater and test–retest reliability for DBD and ADHD symptom scores. Confirmatory factor analysis demonstrated an excellent fit of the DB-DOS multidomain model of DBD symptom scores and a satisfactory fit of ADHD symptom scores. The DB-DOS demonstrated good convergent validity, moderate divergent validity, and good clinical validity on a diagnostic group level for DBD and ADHD symptom scores. The Receiver Operating Characteristic curve analyses revealed that for DBD the sensitivity and specificity are moderate and for ADHD good to excellent. The presumption of a diagnosis based on information from parents, teachers, and cognitive assessment was supported by the DB-DOS in 60% for DBD and 75% for ADHD. The DB-DOS can be used to help support a presumption of a DBD and/or ADHD diagnosis in preschool children.

Likelihood Ratio Tests for Special Rasch Models

In this article, a general class of special Rasch models for dichotomous item scores is considered. Although Andersen’s likelihood ratio test can be used to test whether a Rasch model fits to the data, the test does not differentiate between special Rasch models. Therefore, in this article, new likelihood ratio tests are proposed for testing special Rasch models. The tests proposed do not require individual response pattern frequencies and are useful in practice when the observed total score frequencies are sufficiently large.

Loglinear representations of multivariate Bernoulli Rasch models

In this paper, the extended Rasch model for dichotomously scored items is derived from the general multivariate Bernoulli distribution. The necessary and sufficient conditions for the multivariate Bernoulli distribution to be equal to the extended Rasch model provide a new loglinear representation of the extended Rasch model. Conditions are also given under which the extended Rasch model is equal to the random effects Rasch model, and it is shown under what conditions the extended Rasch model is equal to a random effects Rasch model in which the underlying variable has a normal distribution. In addition, alternative models for the construction of likelihood ratio tests are proposed. One of these alternative models is Habermans extended interaction model. Furthermore, it is shown how both the SPSS and SAS programs can be used to estimate and test loglinear representations of extended Rasch models.

Testing for Two-Way Interactions in the Multigroup Common Factor Model

In this article, a 2-way multigroup common factor model (MG-CFM) is presented. The MG-CFM can be used to estimate interaction effects between 2 grouping variables on 1 or more hypothesized latent variables. For testing the significance of such interactions, a likelihood ratio test is presented. In a simulation study, the robustness of the likelihood ratio test under different sample size conditions is studied and the likelihood ratio test is compared to the Wilkss lambda F test in multivariate analysis of variance (MANOVA) with respect to the approximated power to detect a 2-way interaction effect on a single latent variable. The manipulated factors are the number of indicators, the values of factor loadings, the sample size, and the interaction effect size. The results of the simulation study show that the Type I error rate of the likelihood ratio test is satisfactory and that under all conditions, the approximated power of the likelihood ratio test is considerably higher than that of the Wilkss lambda F test in MANOVA.