A Comparison of Confirmatory Factor Analysis and Network Models for Measurement Invariance Assessment When Indicator Residuals are Correlated.

Social science research is heavily dependent on the use of standardized assessments of a variety of phenomena, such as mood, executive functioning, and cognitive ability. An important assumption when using these instruments is that they perform similarly for all members of the population. When this assumption is violated, the validity evidence of the scores is called into question. The standard approach for assessing the factorial invariance of the measures across subgroups within the population involves multiple groups confirmatory factor analysis (MGCFA). CFA models typically, but not always, assume that once the latent structure of the model is accounted for, the residual terms for the observed indicators are uncorrelated (local independence). Commonly, correlated residuals are introduced after a baseline model shows inadequate fit and inspection of modification indices ensues to remedy fit. An alternative procedure for fitting latent variable models that may be useful when local independence does not hold is based on network models. In particular, the residual network model (RNM) offers promise with respect to fitting latent variable models in the absence of local independence via an alternative search procedure. This simulation study compared the performances of MGCFA and RNM for measurement invariance assessment when local independence is violated, and residual covariances are themselves not invariant. Results revealed that RNM had better Type I error control and higher power compared to MGCFA when local independence was absent. Implications of the results for statistical practice are discussed.

Prediction With Mixed Effects Models: A Monte Carlo Simulation Study.

Oftentimes in many fields of the social and natural sciences, data are obtained within a nested structure (e.g., students within schools). To effectively analyze data with such a structure, multilevel models are frequently employed. The present study utilizes a Monte Carlo simulation to compare several novel multilevel classification algorithms across several varied data conditions for the purpose of prediction. Among these models, the panel neural network and Bayesian generalized mixed effects model (multilevel Bayes) consistently yielded the highest prediction accuracy in test data across nearly all data conditions.

Poverty and Covid-19: Rates of Incidence and Deaths in the United States During the First 10 Weeks of the Pandemic.

The Covid-19 pandemic in the winter and spring of 2020 represents a major challenge to the world health care system that has not been seen perhaps since the influenza pandemic in 1918. The virus has spread across the world, claiming lives on all continents with the exception of Antarctica. Since its arrival in the United States, attention has been paid to how Covid-19 cases and deaths have been distributed across varying socioeconomic and ethnic groups. The goal of this study was to examine this issue during the early weeks of the pandemic, with the hope of shedding some light on how the number of cases and the number of deaths were, or were not related to poverty. Results of this study revealed that during the early weeks of the pandemic more disadvantaged counties in the United States had a larger number of confirmed Covid-19 cases, but that over time this trend changed so that by the beginning of April, 2020 more affluent counties had more confirmed cases of the virus. The number of deaths due to Covid-19 were associated with poorer and more urban counties. Discussion of these results focuses on the possibility that testing for the virus was less available in more disadvantaged counties later in the pandemic than was the case earlier, as the result of an overall lack of adequate testing resources across the nation.

Modeling of self-report behavior data using the generalized covariates in a uniform and shifted binomial mixture model: An empirical example and Monte Carlo simulation.

Social scientists routinely collect data using questionnaires and surveys. Items on these instruments frequently involve scales with multiple ordered options that respondents use to report intensity of feelings or behaviors. Given their popularity, a variety of statistical models have been developed for analyzing data collected using these items. A model that has been recently described for working with ordinal items is the covariates in a uniform and shifted binomial mixture (CUB). The CUB model characterizes responses to ordinal items as a function of two parameters: (a) response feeling (or intensity), and (b) response uncertainty. This model has been extended to include a third parameter measuring likelihood of respondents selecting a socially desirable or safe response, known as the shelter option. This model has been primarily used to investigate items measuring political opinions or product preferences. However, the CUB with a shelter parameter and covariates generalized covariates in a uniform and shifted ninomial mixture model (GeCUB) seems particularly well suited for characterizing self-reported behaviors, particularly those that are not considered positive (i.e., substance abuse). The purpose of this study is to apply this extension of the CUB to the modeling of self-reported substance use behavior by teenagers. Results from the GeCUB model estimation revealed that subjects used the "no use" response as a shelter option at relatively high rates for marijuana use but not for cigarettes or alcohol. In addition, females reported less use and less certainty in their responses than did males. (PsycINFO Database Record (c) 2020 APA, all rights reserved).

Detecting Feigning in Adolescents on the Personality Assessment Inventory-Adolescent Form.

Much of the research on identifying feigning in psychological assessment has focused on adults with less attention to adolescents. The purpose of the present study is to expand the limited literature on detecting feigning in adolescents using the Personality Assessment Inventory - Adolescent. The study included 114 nonclinical adolescents (ages 15 to 18) recruited from high schools in the Midwest who were randomly assigned to experimental groups: honest nonclinical, uncoached feigning, and coached feigning. 50 randomly selected individuals with depression from the PAI-A clinical standardization sample were included as the honest clinical group. Sample demographics included a mean age of 16.64 years; 51.2% young men, 48.2% young women; 85.4% Caucasian, 6.7% African American, 5.5% Hispanic, and 2.4% Asian. 80% of feigning profiles reported clinical levels of depression. MANOVA results showed strong support for the Rogers Discriminant Function (RDF; d range = 1.85-2.05). The Negative Impression Management (NIM) scale also demonstrated promise (d range = 0.77-1.08), while the smallest effects for detecting feigning were found for the Malingering Index (d range = 0.58-0.70). The negative distortion indices showed good utility in differentiating between groups. Cut-scores and pragmatic implications are presented.

Quantifying Item Invariance for the Selection of the Least Biased Assessment.

An important aspect of educational and psychological measurement and evaluation of individuals is the selection of scales with appropriate evidence of reliability and validity for inferences and uses of the scores for the population of interest. One aspect of validity is the degree to which a scale fairly assesses the construct(s) of interest for members of different subgroups within the population. Typically, this issue is addressed statistically through assessment of differential item functioning (DIF) of individual items, or differential bundle functioning (DBF) of sets of items. When selecting an assessment to use for a given application (e.g., measuring intelligence), or which form of an assessment to use in a given instance, researchers need to consider the extent to which the scales work with all members of the population. Little research has examined methods for comparing the amount or magnitude of DIF/DBF present in two assessments when deciding which assessment to use. The current simulation study examines 6 different statistics for this purpose. Results show that a method based on the random effects item response theory model may be optimal for instrument comparisons, particularly when the assessments being compared are not of the same length.

Group membership prediction when known groups consist of unknown subgroups: a Monte Carlo comparison of methods.

Classification using standard statistical methods such as linear discriminant analysis (LDA) or logistic regression (LR) presume knowledge of group membership prior to the development of an algorithm for prediction. However, in many real world applications members of the same nominal group, might in fact come from different subpopulations on the underlying construct. For example, individuals diagnosed with depression will not all have the same levels of this disorder, though for the purposes of LDA or LR they will be treated in the same manner. The goal of this simulation study was to examine the performance of several methods for group classification in the case where within group membership was not homogeneous. For example, suppose there are 3 known groups but within each group two unknown classes. Several approaches were compared, including LDA, LR, classification and regression trees (CART), generalized additive models (GAM), and mixture discriminant analysis (MIXDA). Results of the study indicated that CART and mixture discriminant analysis were the most effective tools for situations in which known groups were not homogeneous, whereas LDA, LR, and GAM had the highest rates of misclassification. Implications of these results for theory and practice are discussed.

Supervised classification in the presence of misclassified training data: a Monte Carlo simulation study in the three group case.

Statistical classification of phenomena into observed groups is very common in the social and behavioral sciences. Statistical classification methods, however, are affected by the characteristics of the data under study. Statistical classification can be further complicated by initial misclassification of the observed groups. The purpose of this study is to investigate the impact of initial training data misclassification on several statistical classification and data mining techniques. Misclassification conditions in the three group case will be simulated and results will be presented in terms of overall as well as subgroup classification accuracy. Results show decreased classification accuracy as sample size, group separation and group size ratio decrease and as misclassification percentage increases with random forests demonstrating the highest accuracy across conditions.