[show abstract][hide abstract] ABSTRACT: The conceptual and methodological framework for measurement equivalence procedures has been well established and widely used. Although multilevel theories and methods have been widely used in organizational research, there is no comparable framework for measurement equivalence of multilevel constructs, or psychometric isomorphism. In this paper, we present a conceptual and methodological framework for understanding and testing various forms of isomorphism. Within this framework, we explicate (a) the different types of psychometric isomorphism, (b) the conditions where psychometric isomorphism is appropriate and necessary, (c) how psychometric isomorphism corresponds with different composition models and estimation methods, and (d) the analytic procedures that can be used. Using simulated data, we also illustrate how the proposed procedures may be applied via two analytic methods – item response theory and factor analysis. We conclude with a discussion of theoretical and methodological implications provided by the proposed framework of psychometric isomorphism.
Organizational Research Methods 01/2014; in press. · 3.26 Impact Factor
[show abstract][hide abstract] ABSTRACT: Prototypes (social images) have been shown to influence behaviour, which is likely to depend on the type of image. Prototype evaluation is based on (un)desirable characteristics related to that image. By an elicitation procedure we examined which adjectives are attributed to specific drinker prototypes. In total 149 young Dutch adults (18-25 years of age) provided adjectives for five drinker prototypes: abstainer, moderate drinker, heavy drinker, tipsy, and drunk person. Twenty-three unique adjectives were found. Multilevel latent class cluster analysis revealed six adjective clusters, each with unique and minor overlapping adjectives: 'negative, excessive drinker,' 'moderate, responsible drinker,' 'funny tipsy drinker,' 'determined abstainer cluster,' 'uncontrolled excessive drinker,' and 'elated tipsy cluster.' In addition, four respondent classes were identified. Respondent classes showed differences in their focus on specific adjective clusters. Classes could be labelled 'focus-on-control class,' 'focus-on-hedonism class,' 'contrasting-extremes-prototypes class,' and 'focus-on-elation class.' Respondent classes differed in gender, educational level and drinking behaviour. The results underscore the importance to differentiate between various prototypes and in prototype adjectives among young adults: subgroup differences in prototype salience and relevance are possibly due to differences in adjective labelling. The results provide insights into explaining differences in drinking behaviour and could potentially be used to target and tailor interventions aimed at lowering alcohol consumption among young adults via prototype alteration.
British Journal of Psychology 08/2013; 104(3):382-399. · 2.37 Impact Factor
[show abstract][hide abstract] ABSTRACT: Workplace bullying has often been attributed to work-related stress, and has been linked to the Job Demand Control Model. The current study aims to further these studies by testing the model for bullying in a heterogeneous sample and by using latent class (LC)-analyses to define different demands and control groups and targets of severe bullying. High job demands were associated with a higher probability of being a target of severe bullying, which was particularly true for the very high job demands group. Low job control was also associated with a higher probability of being a target of severe bullying. Moreover, high job control buffered the negative effects of job demands on being a target of severe bullying, particularly when employees reported very little job control and high/very high job demands. Overall, the JDC-Model was supported, suggesting that being a target of severe bullying can be considered as a social behavioural strain.
Economic and Industrial Democracy 02/2013; 34(1):69-87. · 0.60 Impact Factor
[show abstract][hide abstract] ABSTRACT: In this study we investigated whether we could distinguish the use of specific verbal and visual short term memory (STM) processes in children, or whether the differences in memory performance could be interpreted only in terms of quantitative differences. First, the number of processes involved in the responses on six STM tasks (serial order reconstruction) of 210 primary school children aged 5-12 years was examined by means of latent states. The number of items to reconstruct was manipulated to unravel quantitative differences in responses (high or low performance), and the similarity of the items was manipulated to distinguish qualitative differences in responses (verbal or visual processing). Furthermore, we examined how children changed from one type of process to another on tasks with list lengths of 3, 5, and 7 items by means of the dynamics between the latent states using a latent Markov model. The results showed that two latent states representing the use of specific verbal and visual STM processes could be distinguished on all the tasks. Moreover, two latent states showing merely differences in performance were also found. These findings underline the value of latent variable models to unravel differences between as well as within individuals in the use of cognitive processes.
[show abstract][hide abstract] ABSTRACT: In this paper, we use panel data from the UK and Germany to investigate the effect of employer changes and in-firm job changes on year-to-year wage mobility of male full-time workers. Following segmentation theories and the job search theory, we study whether this effect differs for the low- and high-wage workers. As wage growth is endogenous to the decision of changing jobs, a two-stage Heckman selection approach is used. Specifically, we first estimate a random-effects multinomial logit model for the selection into a job transition and then a fixed-effects panel regression model for the wage growth. The findings suggest that both external and in-firm job changes result into substantial wage gains for the low-paid workers but not for the medium- or high-paid workers. However, the wage gain of low-paid workers due to an in-firm job change is only observed in the UK and is less pronounced than their gain by an external job change. In the German labour market, the later effect is insignificant. The results indicate that low-paid workers profit more from a voluntary change of employer in the coordinated German labour market and from a voluntary in-firm change in the liberal British labour market.
[show abstract][hide abstract] ABSTRACT: We studied four methods for handling incomplete categorical data in statistical modeling: (1) maximum likelihood estimation of the statistical model with incomplete data, (2) multiple imputation using a loglinear model, (3) multiple imputation using a latent class model, (4) and multivariate imputation by chained equations. Each method has advantages and disadvantages, and it is unknown which method should be recommended to practitioners. We reviewed the merits of each method and investigated their effect on the bias and stability of parameter estimates and bias of the standard errors. We found that multiple imputation using a latent class model with many latent classes was the most promising method for handling incomplete categorical data, especially when the number of variables used in the imputation model is large.
Statistical Methods in Medical Research 11/2012; · 2.36 Impact Factor
[show abstract][hide abstract] ABSTRACT: BACKGROUND: Schizophrenia is a complex psychiatric disorder characterized by high phenotypic heterogeneity. Previous studies have distinguished between familial and sporadic forms of schizophrenia and have suggested clinical differentiation between patients and relatives from sporadic and multiplex families. We will introduce a more refined method to distinguish between family subtypes based on psychosis dimension profiles in the relatives of schizophrenia patients. METHODS: Positive, negative, disorganization, mania, and depression scores were assessed in 1,392 relatives. Mixed Model Latent Class Analysis was used to identify family subtypes. A family subtype is a relatively homogeneous group of families with similar symptom profiles in the relatives in these families. Next, we investigated in 616 schizophrenia patients whether family subtype was associated with symptom profiles, IQ, cannabis dependence/abuse, or age of onset of psychosis. RESULTS: Based on the data of relatives, we identified two different family types: "healthy" and "at risk for psychiatric disorder". Patients from at risk families obtained higher positive scores compared to patients from healthy families (Wald(1) = 6.6293, p = 0.010). No significant differences were found in any of the remaining variables. CONCLUSIONS: Our findings confirm the existence of high-risk families and although we did not establish an etiological basis for the distinction between family types, genetic studies might reveal whether family subtype is associated with genetic heterogeneity.
[show abstract][hide abstract] ABSTRACT: Factor analysis is a statistical method for describing the associations among sets of observed variables in terms of a small number of underlying continuous latent variables. Various authors have proposed multilevel extensions of the factor model for the analysis of data sets with a hierarchical structure. These Multilevel Factor Models (MFMs) have in common that—as in multilevel regression analysis—variation at the higher level is modeled using continuous random effects. In this article, we present an alternative multilevel extension of factor analysis which we call the Multilevel Mixture Factor Model (MMFM). It is based on the assumption that higher level units belong to latent classes that differ in terms of the parameters of the factor model specified for the lower level units. We demonstrate the added value of MMFM compared with MFM, both from a theoretical and applied perspective, and we illustrate the complementarity of the two approaches with an empirical application on students' satisfaction with the University of Florence. The multilevel aspect of this application is that students are nested within study programs, which makes it possible to cluster these programs based on their differences in students' satisfaction.
Multivariate Behavioral Research - MULTIVARIATE BEHAV RES. 01/2012; 47(2):247-275.
[show abstract][hide abstract] ABSTRACT: We discuss the limitations of hypothesis testing using (quasi-) experiments in the study of cognitive development and suggest latent variable modeling as a viable alternative to experimentation. Latent variable models allow testing a theory as a whole, incorporating individual differences with respect to developmental processes or abilities in the model. Experiments, in contrast, aim at testing hypotheses that refer to a specific part of a theory; also they ignore individual differences or model the individual differences using age group as a proxy for developmental stage. Drawing on a sample of 409 5–13-year olds, we demonstrate the advantages of latent variable models in the area of transitive reasoning. A comparison of three models showed that the latent variable model that represented fuzzy trace theory had a better fit than the models representing Piaget's theory or linear ordering theory.
[show abstract][hide abstract] ABSTRACT: Fatigue is frequently reported in sarcoidosis and appears to differ between patients. Three types of fatigue (Early Morning Fatigue, Intermittent Fatigue, and Afternoon Fatigue) are described in the literature for sarcoidosis, but have not been validated. Therefore, the aim of this study was to examine whether these types of fatigue can be identified in sarcoidosis.
Outpatients (n=434) from Maastricht University Medical Centre participated in this study. Data were obtained from medical records. Patients also completed questionnaires regarding depressive symptoms, fatigue, quality of life, restless legs, dyspnea, depressive symptoms, anxiety, sleeping problems, symptoms indicative for small fiber neuropathy, and employment.
Latent Cluster Analysis revealed three clusters: 1) Mild Fatigue: patients with mild or no complaints of fatigue, 2) Intermittent Fatigue: patients with complaints of fatigue that varied during the day, and 3) All Day Fatigue: patients who felt tired the whole day. The three patient clusters differed regarding clinical, psychological, and demographical characteristics, with All Day Fatigue patients reporting the most complaints.
Intermittent fatigue was validated and two other types were found. Careful consideration to categorize patients with sarcoidosis in the three types of fatigue will help healthcare providers to understand the challenges these patients encounter. The usefulness of psychological counseling should be evaluated in future research in order to improve the wellbeing of the patients, especially for those with All Day Fatigue.
Journal of psychosomatic research 12/2011; 71(6):416-22. · 2.91 Impact Factor
[show abstract][hide abstract] ABSTRACT: Working memory (WM) processing in children has been studied with different approaches, focusing on either the organizational structure of WM processing during development (factor analytic) or the influence of different task conditions on WM processing (experimental). The current study combined both approaches, aiming to distinguish verbal and visual processing in order to investigate WM development. We investigated recall performance under different task conditions in a sample of 5- to 13-year-olds, applying latent class regression analysis. In this analysis, we examined latent classes (subgroups) within the sample that differed in terms of processing type. The interpretations of the latent classes were validated internally using characteristics of the latent classes and externally using recall performance of words and figures. The results showed that children of different developmental stages used the same type of processing under the same conditions. However, due to developmental differences, their overall performances differed, showing groups of children who were successful in verbal or visual processing and groups of children who were not. This study shows and discusses the importance of disentangling the influence of task conditions from the influence of WM development when interpreting recall performance in children.
[show abstract][hide abstract] ABSTRACT: Dimensional approaches assume that all individuals within hierarchical units (e.g., organizations or countries) share the same measurement model. However, such models are less applicable when researchers are interested in obtaining classes of individuals who share the same measurement model across hierarchical units and to obtain hierarchical latent classes (LCs). The authors present the multilevel mixed-measurement item response theory (MMM-IRT) model as an alternative. This model yields classes of individuals with a common measurement model that spans across hierarchical units. In addition, hierarchical units are classified together to the extent that they share similar proportions of individual-level classes. The authors illustrate the MMM-IRT model with data on self-reported emotions from 121,740 individuals across 116 countries where four individual classes and five country classes were found. Theoretical and methodological implications concerning cross-cultural, multilevel, and measurement equivalence research are discussed.
Organizational Research Methods 08/2011; 14:177-207. · 3.26 Impact Factor
[show abstract][hide abstract] ABSTRACT: Cross-cultural comparison of attitudes using rating scales may be seriously biased by response styles. This paper deals with statistical methods for detection of and correction for extreme response style (ERS), which is one of the well-documented response styles. After providing an overview of available statistical methods for dealing with ERS, we argue that the latent class factor analysis (LCFA) approach proposed by Moors (2003) has several advantages compared to other methods. Moors’ method involves defining a latent variable model which, in addition to the substantive factors of interest, contains an ERS factor. In LCFA the observed ratings can be treated as nominal responses, which is necessary for modeling ERS. We find strong evidence for the presence of ERS and, moreover, find that the groups differ not only in their attitudes but also in ERS. These findings underscore the importance of controlling for ERS when examining attitudes in cross-cultural research.
[show abstract][hide abstract] ABSTRACT: Steinley and Brusco (2011) presented the results of a huge simulation study aimed at evaluating cluster recovery of mixture model clustering (MMC) both for the situation where the number of clusters is known and is unknown. They derived rather strong conclusions on the basis of this study, especially with regard to the good performance of K-means (KM) compared with MMC. I agree with the authors' conclusion that the performance of KM may be equal to MMC in certain situations, which are primarily the situations investigated by Steinley and Brusco. However, a weakness of the paper is the failure to investigate many important real-world situations where theory suggests that MMC should outperform KM. This article elaborates on the KM-MMC comparison in terms of cluster recovery and provides some additional simulation results that show that KM may be much worse than MMC. Moreover, I show that KM is equivalent to a restricted mixture model estimated by maximizing the classification likelihood and comment on Steinley and Brusco's recommendation regarding the use of mixture models for clustering.
[show abstract][hide abstract] ABSTRACT: In this study, we first explore whether different exposure groups of workplace bullying exist, employing a large, heterogeneous sample. The results show six different exposure groups: almost 30.5% is not bullied since they report hardly any negative act at work at all, 27.2% face some limited work criticism, 20.8% face limited negative encounters, 8.3% is occasionally bullied, 9.5% are predominately work related bullied, and a total of 3.6% can be seen victims of severe workplace bullying. In a second step, the relationship between the identified target groups and social demographics were investigated using multinomial logistic regression to identify risk groups of workplace bullying. Employees between the age of 35 and 54, public servants, blue-collar workers, as well as employees working in the food and manufacturing industries have a significantly elevated risk to be victims of workplace bullying. In contrast, employees younger than 25, employees with a temporary contract, teachers, nurses and assistant nurses are those least likely at risk. These findings are important for policymakers at the national and organisational level as they assist in focussing towards possible avenues to prevent workplace bullying.
Industrial Health 01/2011; 49(1):73-88. · 0.87 Impact Factor
[show abstract][hide abstract] ABSTRACT: Traditional item response theory (IRT) measurement invariance approaches examine measurement equivalence (ME) between observed groups (e.g., race, gender, culture). By contrast, mixed-measurement item response theory (MM-IRT) ascertains ME among unobserved groups (i.e., latent classes [LC] of respondents distinguished by differences in scale use). Both approaches can be integrated by using the Mixed-Measurement Item Response Theory with Covariates (MM-IRT-C) model, in which covariates (i.e., observed characteristics) are modeled in conjunction with LCs, thereby elucidating if ME is attributable to observed and/or unobserved groupings. We first show how this technique can be used to ascertain ME over multiple observed characteristics (categorical and/or continuous) concomitantly, thereby advancing a more general approach to observed ME. Next, we illustrate how the full MM-IRT-C can be used to: (a) infer underlying latent measurement classes (LCs), (b) determine associations of LC membership with observed characteristics, and (c) determine if observed measurement nonequivalence occurs predominantly within a particular latent measurement class. This method is demonstrated using a measure of union citizenship behavior, with years of work experience and gender as covariates. The proposed framework extends organizational ME research from considering a single question (i.e., Is there ME between categorical observed groups?) to addressing eight, separate questions about observed and unobserved ME. The substantive and methodological contributions of this model for rethinking ME and its use in organizational research are discussed. 147 Keywords measurement invariance, latent class analysis, profile analysis, item response theory, differential item functioning Observed groupings (e.g., race, gender, and culture) have been integral in the analysis of measurement equivalence (ME). Their use is generally driven by theoretical/practical/legal concerns over whether subgroups use the same frame of reference on a measure of interest (Riordan & Vandenberg, 1994; Vandenberg & Lance, 2000) and whether scores on the measure are comparable across groups (Drasgow, 1987; Stark, Chernyshenko, & Drasgow, 2004). However, focusing on differences in scale use across observed groups is not particularly informative regarding latent individual differences in scale use exist (e.g., response sets/styles, see Eid & Rauber, 2000). In con-trast to the observed ME approach, mixed-measurement item response theory (MM-IRT; Mislevy & Verhelst, 1990; Rost, 1990, 1991) focuses on unobserved ME by identifying latent classes (LCs) of individuals who use scale items in a distinct manner when responding to psychological measures (e.g., Hernandez, Drasgow, & Gonzalez-Roma, 2004; Zickar, Gibby, & Robie, 2004). In this article, we present the use of MM-IRT with covariates (MM-IRT-C; see also Maij-de Meij, Kelderman, & van der Flier, 2008; Smit, Kelderman, & van der Flier, 1999, 2000) as a method for examining both observed and unobserved ME. Using a latent class measurement model in which observed groupings are simultaneously modeled as covariates, we advance an integrated framework for assessing both observed and unobserved ME in organizational research. At this juncture, we clar-ify some terminology: in the IRT literature, measurement nonequivalence is also referred to as dif-ferential item functioning (DIF; for further elaboration, see Stark, Chernyshenko, & Drasgow, 2006; Vandenberg & Lance, 2000). The conceptual differences between observed and unobserved DIF are delineated in Table 1. This table not only serves to show how different ME procedures detect observed or unobserved DIF but also conveys the key notion that differences in scale use may be a function of both observed and unobserved individual characteristics (e.g., Cohen & Bolt, 2005; De Ayala, Kim, Stapleton, & Dayton, 2002; Maij-de Meij et al., 2008).
Organizational Research Methods 01/2011; 14(1):147-176. · 3.26 Impact Factor
[show abstract][hide abstract] ABSTRACT: Three distinctive methods of assessing measurement equivalence of ordinal items, namely, confirmatory factor analysis, differential item functioning using item response theory, and latent class factor analysis, make different modeling assumptions and adopt different procedures. Simulation data are used to compare the performance of these three approaches in detecting the sources of measurement inequivalence. For this purpose, the authors simulated Likert-type data using two nonlinear models, one with categorical and one with continuous latent variables. Inequivalence was set up in the slope parameters (loadings) as well as in the item intercept parameters in a form resembling agreement and extreme response styles. Results indicate that the item response theory and latent class factor models can relatively accurately detect and locate inequivalence in the intercept and slope parameters both at the scale and the item levels. Confirmatory factor analysis performs well when inequivalence is located in the slope parameters but wrongfully indicates inequivalence in the slope parameters when inequivalence is located in the intercept parameters. Influences of sample size, number of inequivalent items in a scale, and model fit criteria on the performance of the three methods are also analyzed.
Sociological Methods & Research 01/2011; 40(2):279-310. · 1.52 Impact Factor