When can categorical variables be treated as continuous? A comparison of robust continuous and categorical SEM estimation methods under suboptimal conditions

A simulation study compared the performance of robust normal theory maximum likelihood (ML) and robust categorical least squares (cat-LS) methodology for estimating confirmatory factor analysis models with ordinal variables. Data were generated from 2 models with 2-7 categories, 4 sample sizes, 2 latent distributions, and 5 patterns of category thresholds. Results revealed that factor loadings and robust standard errors were generally most accurately estimated using cat-LS, especially with fewer than 5 categories; however, factor correlations and model fit were assessed equally well with ML. Cat-LS was found to be more sensitive to sample size and to violations of the assumption of normality of the underlying continuous variables. Normal theory ML was found to be more sensitive to asymmetric category thresholds and was especially biased when estimating large factor loadings. Accordingly, we recommend cat-LS for data sets containing variables with fewer than 5 categories and ML when there are 5 or more categories, sample size is small, and category thresholds are approximately symmetric. With 6-7 categories, results were similar across methods for many conditions; in these cases, either method is acceptable. (PsycINFO Database Record (c) 2012 APA, all rights reserved).     

Categorical perception of morphed objects using a free-naming experiment

Morphed figures entail a dominant and nondominant interpretation. Testing perception of morphed objects using forced-choice methods demonstrates that morphed figures are perceived as their dominant interpretation (“categorical perception”, or CP). Using a more natural free-naming response could reveal whether CP is an effect independent of method. In Experiment 1, therefore, series of morphed figures were tested for CP using free naming. Half of the morph series were identified as CP patterns. In Experiment 2, we used forced choice to investigate CP, resulting in an increase of number of CP series compared to free naming. The overlap between CP series of Experiments 1 and 2 was small, however. Experiment 3 revealed that higher perceptual similarity between the extremes of the series was strongly related to CP for the free-naming method, in contrast to the forced-choice method. We conclude that the observation of CP depends on the intactness of the intrinsic object structure caused by the morphing procedure.     

The performance of robust test statistics with categorical data

This paper reports on a simulation study that evaluated the performance of five structural equation model test statistics appropriate for categorical data. Both Type I error rate and power were investigated. Different model sizes, sample sizes, numbers of categories, and threshold distributions were considered. Statistics associated with both the diagonally weighted least squares (cat‐DWLS) estimator and with the unweighted least squares (cat‐ULS) estimator were studied. Recent research suggests that cat‐ULS parameter estimates and robust standard errors slightly outperform cat‐DWLS estimates and robust standard errors ( Forero, Maydeu‐Olivares, & Gallardo‐Pujol, 2009 ). The findings of the present research suggest that the mean‐ and variance‐adjusted test statistic associated with the cat‐ULS estimator performs best overall. A new version of this statistic now exists that does not require a degrees‐of‐freedom adjustment ( Asparouhov & Muthén, 2010 ), and this statistic is recommended. Overall, the cat‐ULS estimator is recommended over cat‐DWLS, particularly in small to medium sample sizes.     

the inventory of recent life experiences for cancer patients (IRLE-C): A decontaminated measure of cancer-based hassles

Abstract

Cancer is recognized to have multifaceted stressful impact on all areas of a patient's life. Researchers commonly use self-report questionnaires, intended to measure stressors objectively. However, the item-content and response-format of such scales often tap physical and mental responses to stress, thereby contaminating prediction of adverse impact. This article reports the development and validation of English and French versions of the Inventory of Recent Life Experiences for Cancer Patients (IRLE-C) which is designed to minimize such “criterion-contamination”. This entailed (1) avoiding items reflecting physical or subjective distress; (2) rating stressors for degree of exposure only; and (3) use of an innocuous scale title. The initial item pool was administered serially to a sample of 100 Francophone breast-cancer and prostate-cancer patients. To guard against inflating reliability and validity estimates through capitalizing on chance, we administered the 30-item final scale to an independent sample of 96 Francophone breast-cancer and prostate-cancer patients undergoing radiation treatment. Following the item-selection step, factorial structure and validity analyses were performed using the combined French-speaking sample (n= 196). Second, we administered the English version of the scale to an English-speaking sample of 127 cancer patients (various cancer sites and stages). The measure showed good internal consistency (.94 and .89 for the Francophone and Anglophone samples respectively) and met criteria for a 2-week test-retest reliability (r= .70 for the item-selection subsample and .80 for the cross-replication sub-sample). Correlations between the IRLE-C and the POMS Total Mood Disturbance were around .60 for both the Francophone and Anglophone samples. Avoiding contamination (through content and format) without losing its relationship to subjective distress, the IRLE-C appears a useful instrument for applying the stress-process model in oncology to establish clear distinctions among stressors, mediators, reactions, and consequences.     

Reexamining the possible benefits of visual crowding: dissociating crowding from ensemble percepts

Peripheral objects and their features become indistinct when closely surrounding but nonoverlapping objects are present. Most models suggest that this phenomenon, called crowding, reflects limitations of visual processing, but an intriguing idea is that it may be, in part, adaptive. Specifically, the mechanism generating crowding may simultaneously facilitate ensemble representations of features, leaving meaningful information about clusters of objects. In two experiments, we tested whether visual crowding and the perception of ensemble features share a common mechanism. Observers judged the orientation of a crowded bar, or the ensemble orientation of all bars in the upper and lower visual fields. While crowding was predictably stronger in the upper relative to the lower visual field, the ensemble percept did not vary between the visual fields. Featural averaging within the crowded region does not always scale with the resolution limit defined by crowding, suggesting that dissociable processes contribute to visual crowding and ensemble percepts.     

The moderating influence of procedural fairness on the relationship between work-life conflict and organizational commitment

To help employees better manage work-life conflict, organizations have introduced various initiatives, which have met with mixed results. The present studies examined the utility of a procedurally based approach to understanding employees' reactions to work-life conflict. The authors examined whether the fairness of procedures used by organizational authorities to plan and implement decisions moderates the (inverse) relationship between work-life conflict and employees' organizational commitment. Three studies using different methodologies showed support for the moderating role played by procedural fairness. That is, the tendency for greater work-life conflict to lead to lower commitment was significantly less pronounced when procedural fairness was high rather than low. Theoretical contributions to the work-life conflict and organizational justice literatures are discussed, as are practical implications.     

Using genetic programming to discover nonlinear variable interactions

Psychology has to deal with many interacting variables. The analyses usually used to uncover such relationships have many constraints that limit their utility. We briefly discuss these and describe recent work that uses genetic programming to evolve equations to combine variables in nonlinear ways in a number of different domains. We focus on four studies of interactions from lexical access experiments and psychometric problems. In all cases, genetic programming described nonlinear combinations of items in a manner that was subsequently independently verified. We discuss the general implications of genetic programming and related computational methods for multivariate problems in psychology.