Articles

See also my Google Scholar profile.

Showing only journal articles. Click here to show all publications.

submitted

  • {Van} Erp, S., Mulder, J., & Oberski, D. L.. (submitted). Prior sensitivity analysis in default bayesian structural equation modeling. .
    [BibTeX] [Download PDF]
    @article{vanerp2016sensitivity,
    Author = {{Van} Erp, Sara and Mulder, Joris and D. L. Oberski},
    Date-Added = {2016-04-29 17:24:18 +0000},
    Date-Modified = {2016-04-29 17:25:07 +0000},
    Title = {Prior Sensitivity Analysis in Default Bayesian Structural Equation Modeling},
    Year = {submitted}}

  • Oberski, D. L.. (submitted). Model-based variance estimation for aggregated covariance structure models. .
    [BibTeX] [Abstract] [Download PDF]
    Covariance structure models comprise a wide class of models, popular in the social and behavioral sciences and often applied to complex sample surveys. Variance estimation for such models has received relatively little attention. This short note introduces a model-based variance estimator under complex sampling for aggregated parameters of covariance structure models. This variance estimator can be used for three purposes: to assess sampling variance when the model is thought to be correct; as a working covariance matrix in GEE estimation; or to estimate "design" (or "misspecification") effects of nonnormality and clustering separately. A small simulation study indicates that the proposed estimator can accurately recover sampling variance, while an example confirmatory factor analysis demonstrates its use. Key words: Structural equation modeling, covariance structure modeling, complex sampling, variance estimation, design effects.
    @article{Oberski:WP:complex-NT,
    Abstract = {
    Covariance structure models comprise a wide class of models, popular in the social and behavioral sciences and often applied to complex sample surveys. Variance estimation for such models has received relatively little attention. This short note introduces a model-based variance estimator under complex sampling for aggregated parameters of covariance structure models. This variance estimator can be used for three purposes: to assess sampling variance when the model is thought to be correct; as a working covariance matrix in GEE estimation; or to estimate "design" (or "misspecification") effects of nonnormality and clustering separately. A small simulation study indicates that the proposed estimator can accurately recover sampling variance, while an example confirmatory factor analysis demonstrates its use.
    Key words: Structural equation modeling, covariance structure modeling, complex sampling, variance estimation, design effects.
    },
    Author = {Oberski, D. L.},
    Date-Added = {2013-08-07 07:37:12 +0000},
    Date-Modified = {2013-08-07 07:38:38 +0000},
    Title = {Model-based variance estimation for aggregated covariance structure models},
    Url = {http://daob.nl/wp-content/uploads/2013/08/Model-based-variance-estimation-for-aggregated-covariance-structure-models.pdf},
    Year = {submitted},
    Bdsk-Url-1 = {http://daob.nl/wp-content/uploads/2013/08/Model-based-variance-estimation-for-aggregated-covariance-structure-models.pdf}}

  • Oberski, D. L.. (submitted). A flexible method to explain differences in structural equation model parameters over subgroups. .
    [BibTeX] [Abstract] [Download PDF] [Download data and code from the paper]
    Structural Equation Modeling (SEM) is a widely applied technique to model relationships including both observed and latent variables. Recently, the explanation of subgroup differences in SEM parameters has attracted the attention of applied social researchers. Examples are differences in survey measurement error variance and differences in the variance of, possibly only indirectly observed, wages or educational achievement to explain inequality. However, these fields have been hindered by the lack of methods to regress a SEM parameter such as a latent variable's variance on a set of predictor variables. I present a method, "IPC regression", addressing this problem. IPC regression involves three analysis steps: 1) running a pooled SEM, 2) calculating so-called "individual parameter contributions" (IPC's), and 3) regressing these IPC's on the predictors of scientific interest. IPC regression is presented in a general framework that encompasses any type of SEM parameter and prediction model. Largeand small-sample properties are considered, and an example application illustrates the implications for practical research.
    @article{Oberski:WP:SEM-IPC,
    Abstract = {
    Structural Equation Modeling (SEM) is a widely applied technique to model relationships including both observed and latent variables. Recently, the explanation of subgroup differences in SEM parameters has attracted the attention of applied social researchers. Examples are differences in survey measurement error variance and differences in the variance of, possibly only indirectly observed, wages or educational achievement to explain inequality. However, these fields have been hindered by the lack of methods to regress a SEM parameter such as a latent variable's variance on a set of predictor variables. I present a method, "IPC regression", addressing this problem. IPC regression involves three analysis steps: 1) running a pooled SEM, 2) calculating so-called "individual parameter contributions" (IPC's), and 3) regressing these IPC's on the predictors of scientific interest. IPC regression is presented in a general framework that encompasses any type of SEM parameter and prediction model. Largeand small-sample properties are considered, and an example application illustrates the implications for practical research.
    },
    Author = {Oberski, D. L.},
    Datapackage = {http://goo.gl/fstgW},
    Date-Modified = {2013-06-30 11:33:51 +0000},
    Title = {A Flexible Method to Explain Differences in Structural Equation Model Parameters over Subgroups},
    Url = {http://daob.nl/wp-content/uploads/2013/06/SEM-IPC-manuscript-new.pdf},
    Year = {submitted},
    Bdsk-Url-1 = {http://daob.nl/wp-content/uploads/2013/06/SEM-IPC-manuscript-new.pdf},
    Bdsk-Url-2 = {http://daob.nl/wp-content/uploads/2013/06/SEM-IPC-manuscript-new.pdf}}

  • Gallego, A., & Oberski, D. L.. (submitted). Civic duty and voter turnout in the household. .
    [BibTeX] [Download PDF]
    @article{gallego:WP:duty,
    Author = {Gallego, A. and Oberski, D. L.},
    Title = {Civic Duty and Voter Turnout in the Household},
    Url = {http://daob.nl/wp-content/uploads/2013/03/Duty-Household.pdf},
    Year = {submitted},
    Bdsk-Url-1 = {http://daob.nl/wp-content/uploads/2013/03/Duty-Household.pdf}}

  • Pankowska, P., Bakker, B., Pavlopoulos, D., & Oberski, D. L.. (submitted). Reconciliation of two data sources by correction for measurement error: a feasibility study. .
    [BibTeX]
    @Article{pankowska:WP:reconciliation,
    author = {Pankowska, P. and Bakker, B. and Pavlopoulos, D. and Oberski, D. L.},
    title = {Reconciliation of two data sources by correction for measurement error: a feasibility study},
    year = {submitted},
    }

conditionally accepted

  • Oberski, D. L., & Vermunt, J. K.. (conditionally accepted). The expected parameter change (EPC) for local dependence assessment in binary data latent class models. Psychometrika.
    [BibTeX] [Abstract] [Download PDF] [Download data and code from the paper]
    Binary data latent class models crucially assume local independence, violations of which can seriously bias the results. We present two tools for monitoring local dependence in binary data latent class models: the "Expected Parameter Change" (EPC) and a generalized EPC, estimating the substantive size and direction of possible local dependencies. The asymptotic and finite sample behavior of the measures is studied, and two applications to the U.S. Census estimation of Hispanic ethnicity and medical experts' ratings of x-rays demonstrate its value in arriving at a model that balances realism and parsimony. R code implementing our proposal and including both example datasets is available online as supplementary material.
    @article{Oberski:LCA-EPC-pmet,
    Abstract = {
    Binary data latent class models crucially assume local independence, violations of which can seriously bias the results. We present two tools for monitoring local dependence in binary data latent class models: the "Expected Parameter Change" (EPC) and a generalized EPC, estimating the substantive size and direction of possible local dependencies. The asymptotic and finite sample behavior of the measures is studied, and two applications to the U.S. Census estimation of Hispanic ethnicity and medical experts' ratings of x-rays demonstrate its value in arriving at a model that balances realism and parsimony.
    R code implementing our proposal and including both example datasets is available online as supplementary material.
    },
    Author = {Oberski, D. L. and Vermunt, J.K.},
    Datapackage = {http://daob.nl/wp-content/uploads/2013/03/LCA-EPC-R-code-supplement.zip},
    Date-Added = {2014-03-14 17:59:08 +0000},
    Date-Modified = {2014-03-14 17:59:08 +0000},
    Journal = {Psychometrika},
    Title = {The Expected Parameter Change ({EPC}) for Local Dependence Assessment in Binary Data Latent Class Models},
    Url = {http://daob.nl/wp-content/uploads/2014/07/lca-epc-revision2.pdf},
    Year = {conditionally accepted},
    Bdsk-Url-1 = {http://daob.nl/wp-content/uploads/2014/07/lca-epc-revision2.pdf}}

  • Boeschoten, L., Oberski, D. L., & Waal, T. D.. (conditionally accepted). Estimating classification error under edit restrictions in combined survey-register data using multiple imputation latent class modelling (milc). Journal of official statistics.
    [BibTeX] [Abstract] [Download PDF]
    (Note: this is the published article version of the CBS report with the same title and authors)
    @Article{boeschoten2017milc,
    author = {Boeschoten, L. and Oberski, D. L. and Ton De Waal},
    title = {Estimating classification error under edit restrictions in combined survey-register data using Multiple Imputation Latent Class modelling (MILC)},
    journal = {Journal of Official Statistics},
    year = {conditionally accepted},
    abstract = {(Note: this is the published article version of the CBS report with the same title and authors)},
    }

accepted

  • Borghuis, J., Denissen, J. J. A., Oberski, D. L., Sijtsma, K., Meeus, W. H. J., Branje, S., Koot, H. M., & Bleidorn, W.. (accepted). Big five personality (co-)development in adolescent friendship and sibling dyads: a 13-wave longitudinal study. Journal of personality and social psychology. doi:10.1037/pspp0000138
    [BibTeX]
    @Article{borghuis2016bigfive,
    author = {J. Borghuis and J.J.A. Denissen and D.L. Oberski and K. Sijtsma and W.H.J. Meeus and S. Branje and H.M. Koot and W. Bleidorn},
    title = {Big Five Personality (Co-)Development in Adolescent Friendship and Sibling Dyads: A 13-Wave Longitudinal Study},
    journal = {Journal of Personality and Social Psychology},
    doi = {10.1037/pspp0000138},
    year = {accepted},
    }

  • Oberski, D. L., Kirchner, A., Eckman, S., & Kreuter, F.. (accepted). Evaluating the quality of survey and administrative data with generalized multitrait-multimethod models. Journal of the American Statistical Association. doi:10.1080/01621459.2017.1302338
    [BibTeX] [Abstract] [Download PDF] [Download data and code from the paper]
    Administrative register data are increasingly important in statistics, but, like other types of data, may contain measurement errors. To prevent such errors from invalidating analyses of scientific interest, it is therefore essential to estimate the extent of measurement errors in administrative data. Currently, however, most approaches to evaluate such errors involve either prohibitively expensive audits or comparison with a survey that is assumed perfect. We introduce the "generalized multitrait-multimethod'' (GMTMM) model, which can be seen as a general framework for evaluating the quality of administrative and survey data simultaneously. This framework allows both survey and register to contain random and systematic measurement errors. Moreover, it accommodates common features of administrative data such as discreteness, nonlinearity, and nonnormality, improving similar existing models. The use of the GMTMM model is demonstrated by application to linked survey-register data from the German Federal Employment Agency on income from and duration of employment, and a simulation study evaluates the estimates obtained. KEY WORDS: Measurement error, Latent Variable Models, Official statistics, Register data, Reliability
    @Article{oberski2017gmtmm,
    author = {D. L. Oberski and A. Kirchner and S. Eckman and F. Kreuter},
    title = {Evaluating the quality of survey and administrative data with generalized multitrait-multimethod models},
    journal = {{Journal of the American Statistical Association}},
    doi = {10.1080/01621459.2017.1302338},
    year = {accepted},
    abstract = {Administrative register data are increasingly important in statistics, but, like other types of data, may contain measurement errors. To prevent such errors from invalidating analyses of scientific interest, it is therefore essential to estimate the extent of measurement errors in administrative data. Currently, however, most approaches to evaluate such errors involve either prohibitively expensive audits or comparison with a survey that is assumed perfect.
    We introduce the "generalized multitrait-multimethod'' (GMTMM) model, which can be seen as a general framework for evaluating the quality of administrative and survey data simultaneously. This framework allows both survey and register to contain random and systematic measurement errors. Moreover, it accommodates common features of administrative data such as discreteness, nonlinearity, and nonnormality, improving similar existing models. The use of the GMTMM model is demonstrated by application to linked survey-register data from the German Federal Employment Agency on income from and duration of employment, and a simulation study evaluates the estimates obtained.
    KEY WORDS: Measurement error, Latent Variable Models, Official statistics, Register data, Reliability},
    bdsk-url-1 = {http://arxiv.org/abs/1508.05502},
    datapackage = {http://daob.nl/wp-content/uploads/2017/02/oberski-mtmm-supplementary.zip},
    url = {http://arxiv.org/abs/1508.05502},
    }

2017

  • Mayor, J. R., Nathan J. Sanders, Aimée T. Classen, Bardgett, R. D., Clément, J., Fajardo, A., Lavorel, S., Maja K. Sundqvist, Bahn, M., Chisholm, C., Ellen Cieraad, Gedalof, Z., Grigulis, K., Kudo, G., Oberski, D. L., & Wardle, D. A.. (2017). Elevation alters ecosystem properties across temperate treelines globally. Nature, 542. doi:10.1038/nature21027
    [BibTeX]
    @Article{mayor2016elevation,
    author = {Jordan R. Mayor and Nathan J. Sanders, and Aimée T. Classen, and Richard D. Bardgett and Jean-Christophe Clément and Alex Fajardo and Sandra Lavorel and Maja K. Sundqvist, and Michael Bahn and Chelsea Chisholm and Ellen Cieraad, and Ze’ev Gedalof and Karl Grigulis and Gaku Kudo and D.L. Oberski and David A. Wardle},
    title = {Elevation alters ecosystem properties across temperate treelines globally},
    journal = {Nature},
    year = {2017},
    volume = {542},
    issue = {7639},
    doi = {10.1038/nature21027},
    }

2016

  • Di Mari, R., Oberski, D. L., & Vermunt, J. K.. (2016). Bias-adjusted three-step latent Markov modeling with covariates. Structural equation modeling. doi:10.1080/10705511.2016.1191015
    [BibTeX] [Abstract] [Download PDF]
    Latent Markov models with covariates can be estimated via one-step Maximum Likelihood. However, this one-step approach has various disadvantages, such as that the inclusion of covariates in the model may alter the formation of the latent states and that parameter estimation may become infeasible with large numbers of time points, responses, and covariates. This is why researchers typically prefer performing the analysis in a stepwise manner; that is, they first construct the measurement model, then obtain the latent state classifications, and subsequently study the relationship between covariates and latent state memberships. However, such a stepwise approach yields downward biased estimates of the covariate effects on initial state and transition probabilities. The present paper shows how to overcome this problem using a generalization of the bias-corrected three-step estimation method proposed for latent class analysis (Asparouhov & Muthén, 2014; Bolck, Croon, & Hagenaars, 2004; Vermunt, 2010). We give a formal derivation of the generalization to latent Markov models and discuss how it can be used with many time points by incorporating it into a Baum-Welch type of EM algorithm. We evaluate the method through a simulation study and illustrate it using an application on household financial portfolio change. Our study shows that the proposed correction method yields unbiased parameter estimates and accurate standard errors, except for situations with very poorly separated classes and a small sample. Keywords: latent Markov model, hidden Markov model, three-step approach, bias-correction, latent transition analysis, latent class analysis, classify-analyze
    @article{dimari2016bias,
    abstract = {
    Latent Markov models with covariates can be estimated via one-step Maximum Likelihood. However, this one-step approach has various disadvantages, such as that the inclusion of covariates in the model may alter the formation of the latent states and that parameter estimation may become infeasible with large numbers of time points, responses, and covariates. This is why researchers typically prefer performing the analysis in a stepwise manner; that is, they first construct the measurement model, then obtain the latent state classifications, and subsequently study the relationship between covariates and latent state memberships. However, such a stepwise approach yields downward biased estimates of the covariate effects on initial state and transition probabilities. The present paper shows how to overcome this problem using a generalization of the bias-corrected three-step estimation method proposed for latent class analysis (Asparouhov & Muthén, 2014; Bolck, Croon, & Hagenaars, 2004; Vermunt, 2010).
    We give a formal derivation of the generalization to latent Markov models and discuss how it can be used with many time points by incorporating it into a Baum-Welch type of EM algorithm. We evaluate the method through a simulation study and illustrate it using an application on household financial portfolio change. Our study shows that the proposed correction method yields unbiased parameter estimates and accurate standard errors, except for situations with very poorly separated classes and a small sample.
    Keywords: latent Markov model, hidden Markov model, three-step approach, bias-correction, latent transition analysis, latent class analysis, classify-analyze},
    Author = {Di Mari, Roberto and Oberski, D. L. and Vermunt, J.K.},
    Date-Added = {2016-03-03 09:14:34 +0000},
    Date-Modified = {2016-03-03 09:18:17 +0000},
    Title = {Bias-adjusted three-step latent {Markov} modeling with covariates},
    journal = {Structural Equation Modeling},
    doi = {10.1080/10705511.2016.1191015},
    Year = {2016}}

  • Lamont, A. E., Lyons, M. D., Jaki, T., Stuart, E. A., Feaster, D. J., Tharmaratnam, K., Oberski, D. L., Ishwaran, H., Wilson D.K., & Van Horn, M. L.. (2016). Identification of predicted individual treatment effects in randomized clinical trials. Statistical methods in medical research. doi:10.1177/0962280215623981
    [BibTeX] [Abstract]
    In most medical research, treatment effectiveness is assessed using the Average Treatment Effect (ATE) or some version of subgroup analysis. The practice of individualized or precision medicine, however, requires new approaches that predict how an individual will respond to treatment, rather than relying on aggregate measures of effect. In this study, we present a conceptual framework for estimating individual treatment effects, referred to as Predicted Individual Treatment Effects (PITE). We first apply the PITE approach to a randomized controlled trial designed to improve behavioral and physical symptoms. Despite trivial average effects of the intervention, we show substantial heterogeneity in predicted individual treatment response using the PITE approach. The PITEs can be used to predict individuals for whom the intervention may be most effective (or harmful). Next, we conduct a Monte Carlo simulation study to evaluate the accuracy of Predicted Individual Treatment Effects. We compare the performance of two methods used to obtain predictions: multiple imputation and non-parametric random decision trees (RDT). Results showed that, on average, both predictive methods produced accurate estimates at the individual level; however, the RDT tended to underestimate the PITE for people at the extreme and showed more variability in predictions across repetitions compared to the imputation approach. Limitations and future directions are discussed.
    @article{lamont2016pite,
    Abstract = {In most medical research, treatment effectiveness is assessed using the Average Treatment Effect (ATE) or some version of subgroup analysis. The practice of individualized or precision medicine, however, requires new approaches that predict how an individual will respond to treatment, rather than relying on aggregate measures of effect. In this study, we present a conceptual framework for estimating individual treatment effects, referred to as Predicted Individual Treatment Effects (PITE). We first apply the PITE approach to a randomized controlled trial designed to improve behavioral and physical symptoms. Despite trivial average effects of the intervention, we show substantial heterogeneity in predicted individual treatment response using the PITE approach. The PITEs can be used to predict individuals for whom the intervention may be most effective (or harmful). Next, we conduct a Monte Carlo simulation study to evaluate the accuracy of Predicted Individual Treatment Effects. We compare the performance of two methods used to obtain predictions: multiple imputation and non-parametric random decision trees (RDT). Results showed that, on average, both predictive methods produced accurate estimates at the individual level; however, the RDT tended to underestimate the PITE for people at the extreme and showed more variability in predictions across repetitions compared to the imputation approach. Limitations and future directions are discussed.},
    Author = {Lamont, A.E. and Lyons, M.D. and Jaki, T. and Stuart, E.A. and Feaster, D.J. and Tharmaratnam, K. and Oberski, D.L. and Ishwaran, H. and Wilson, D.K., and Van Horn, M.L.},
    Doi = {10.1177/0962280215623981},
    Journal = {Statistical Methods in Medical Research},
    Title = {Identification of predicted individual treatment effects in randomized clinical trials},
    Year = {2016},
    Bdsk-Url-1 = {http://dx.doi.org/10.1177/0962280215623981}}

  • Molenaar, D., Oberski, D. L., Vermunt, J. K., & De Boeck, P.. (2016). Hidden Markov IRT models for responses and response times. Multivariate behavioral research. doi:10.1080/00273171.2016.1192983
    [BibTeX] [Abstract] [Download PDF]
    Current approaches to model responses and response times to psychometric tests assume constant ability and constant speed of the respondent. Violations of this assumption are generally absorbed in the residual of the model. However, explicitly modeling the dynamics of ability and speed can be valuable to detect different solution strategies or different response processes. In this paper we propose a dynamic approach for responses and response times based on hidden Markov modeling. A simulation study is conducted to demonstrate acceptable power and parameter recovery of the model. In addition, both a confirmatory and an exploratory application are presented to demonstrate the practical value of the modeling approach. Keywords: Response time modeling; Hidden Markov modeling; Item response theory; Latent class models; Dynamic modeling; conditional independence.
    @Article{molenaar2016hmm,
    author = {Molenaar, Dylan and Oberski, D. L. and Vermunt, J.K. and De Boeck, P.},
    title = {Hidden {Markov} {IRT} Models for Responses and Response Times},
    journal = {Multivariate Behavioral Research},
    year = {2016},
    abstract = {Current approaches to model responses and response times to psychometric tests assume constant ability and constant speed of the respondent. Violations of this assumption are generally absorbed in the residual of the model. However, explicitly modeling the dynamics of ability and speed can be valuable to detect different solution strategies or different response processes. In this paper we propose a dynamic approach for responses and response times based on hidden Markov modeling. A simulation study is conducted to demonstrate acceptable power and parameter recovery of the model. In addition, both a confirmatory and an exploratory application are presented to demonstrate the practical value of the modeling approach.
    Keywords: Response time modeling; Hidden Markov modeling; Item response theory; Latent class models; Dynamic modeling; conditional independence.},
    date-added = {2015-11-03 14:59:53 +0000},
    date-modified = {2015-11-03 15:02:21 +0000},
    doi = {10.1080/00273171.2016.1192983},
    keywords = {Conditional independence, dynamic modeling, hidden Markov modeling, item response theory, latent class models, response time modeling},
    }

  • Oberski, D. L.. (2016). A review of "latent variable modeling with R". Journal of educational and behavioral statistics, 41, 226-233. doi:10.3102/1076998615621305
    [BibTeX] [Download PDF]
    @article{oberski2016review,
    Author = {Oberski, D. L.},
    Date-Added = {2015-10-27 16:26:39 +0000},
    Date-Modified = {2015-11-09 17:38:44 +0000},
    Doi = {10.3102/1076998615621305},
    Issue = {2},
    Journal = {Journal of Educational and Behavioral Statistics},
    Pages = {226-233},
    Title = {A Review of "Latent Variable Modeling with {R}"},
    Volume = {41},
    Year = {2016},
    Bdsk-Url-1 = {http://dx.doi.org/10.3102/1076998615621305}}

  • Gallego, A., Buscha, F., Sturgis, P., & Oberski, D.. (2016). Places and preferences: a longitudinal analysis of self-selection and contextual effects. British journal of political science, 46, 529-550. doi:10.1017/S0007123414000337
    [BibTeX]
    @article{gallego2014places,
    Author = {Gallego, Aina and Buscha, Franz and Sturgis, Patrick and Oberski, Daniel},
    Doi = {10.1017/S0007123414000337},
    Journal = {British Journal of Political Science},
    Pages = {529-550},
    volume = {46},
    issue = {3},
    Publisher = {Cambridge Univ Press},
    Title = {Places and preferences: A longitudinal analysis of self-selection and contextual effects},
    Year = {2016},
    Bdsk-Url-1 = {http://dx.doi.org/10.1017/S0007123414000337}}

  • {Van} Smeden, M., Oberski, D. L., Vermunt, J. K., Reitsma, H., de Groot, J., & Moons, C.. (2016). Problems in detecting misfit of latent class models in diagnostic research without a gold standard were shown. Journal of clinical epidemiology, 158-66. doi:10.1016/j.jclinepi.2015.11.012
    [BibTeX] [Abstract]
    OBJECTIVES: The objective of this study was to evaluate the performance of goodness-of-fit testing to detect relevant violations of the assumptions underlying the criticized "standard" two-class latent class model. Often used to obtain sensitivity and specificity estimates for diagnostic tests in the absence of a gold reference standard, this model relies on assuming that diagnostic test errors are independent. When this assumption is violated, accuracy estimates may be biased: goodness-of-fit testing is often used to evaluate the assumption and prevent bias. STUDY DESIGN AND SETTING: We investigate the performance of goodness-of-fit testing by Monte Carlo simulation. The simulation scenarios are based on three empirical examples. RESULTS: Goodness-of-fit tests lack power to detect relevant misfit of the standard two-class latent class model at sample sizes that are typically found in empirical diagnostic studies. The goodness-of-fit tests that are based on asymptotic theory are not robust to the sparseness of data. A parametric bootstrap procedure improves the evaluation of goodness of fit in the case of sparse data. CONCLUSION: Our simulation study suggests that relevant violation of the local independence assumption underlying the standard two-class latent class model may remain undetected in empirical diagnostic studies, potentially leading to biased estimates of sensitivity and specificity. KEYWORDS: Goodness of fit; Latent class analysis; Local independence assumption; No gold standard; Sensitivity and specificity; Simulation
    @article{smeden2016detecting,
    Abstract = {
    OBJECTIVES:
    The objective of this study was to evaluate the performance of goodness-of-fit testing to detect relevant violations of the assumptions underlying the criticized "standard" two-class latent class model. Often used to obtain sensitivity and specificity estimates for diagnostic tests in the absence of a gold reference standard, this model relies on assuming that diagnostic test errors are independent. When this assumption is violated, accuracy estimates may be biased: goodness-of-fit testing is often used to evaluate the assumption and prevent bias.
    STUDY DESIGN AND SETTING:
    We investigate the performance of goodness-of-fit testing by Monte Carlo simulation. The simulation scenarios are based on three empirical examples.
    RESULTS:
    Goodness-of-fit tests lack power to detect relevant misfit of the standard two-class latent class model at sample sizes that are typically found in empirical diagnostic studies. The goodness-of-fit tests that are based on asymptotic theory are not robust to the sparseness of data. A parametric bootstrap procedure improves the evaluation of goodness of fit in the case of sparse data.
    CONCLUSION:
    Our simulation study suggests that relevant violation of the local independence assumption underlying the standard two-class latent class model may remain undetected in empirical diagnostic studies, potentially leading to biased estimates of sensitivity and specificity.
    KEYWORDS:
    Goodness of fit; Latent class analysis; Local independence assumption; No gold standard; Sensitivity and specificity; Simulation
    },
    Author = {{Van} Smeden, M. and Oberski, D. L. and Vermunt, J.K. and Reitsma, H. and de Groot, J. and Moons, C.},
    Date-Added = {2015-04-15 14:02:28 +0000},
    Date-Modified = {2015-11-02 13:52:08 +0000},
    Doi = {10.1016/j.jclinepi.2015.11.012},
    Journal = {Journal of Clinical Epidemiology},
    Title = {Problems in detecting misfit of latent class models in diagnostic research without a gold standard were shown},
    Year = {2016},
    issue = {74},
    pages = {158-66},
    Bdsk-Url-1 = {http://dx.doi.org/10.1016/j.jclinepi.2015.11.012}}

  • Bakk, Z., Oberski, D. L., & Vermunt, J. K.. (2016). Relating latent class membership to continuous distal outcomes: improving the LTB approach and a modified three-step implementation. Structural equation modeling: a multidisciplinary journal, 23. doi:10.1080/10705511.2015.1049698
    [BibTeX] [Abstract] [Download PDF]
    The LTB approach relates latent classes (LCs) to distal outcomes by estimating a LC model with the outcome treated as covariate. Based on this model the class-specific means of the outcome are calculated. In this manner no distributional assumptions about the outcome are made Lanza et al. (2013). We provide a stepwise implementation of the approach that separates the building of the latent classes and the investigation of the relationship of the classes with the outcomes. Next, similar to quadratic discriminant analysis, we propose including a quadratic term in the logistic model for the LCs when the variances of the outcome are heteroskedastic in order to prevent parameter bias. And lastly we propose two alternative SE estimators (non-parametric bootstrap, jackknife), that yield better coverage rates than the currently used SE estimator proposed by Asparouhov and Muthén (2014) . The proposed improvements are tested via a simulation study with good results, and applied to real data.
    @article{bakk2015ltb,
    Abstract = {
    The LTB approach relates latent classes (LCs) to distal outcomes by estimating a LC model with the outcome treated as covariate. Based on this model the class-specific means of the outcome are calculated. In this manner no distributional assumptions about the outcome are made Lanza et al. (2013).
    We provide a stepwise implementation of the approach that separates the building of the latent classes and the investigation of the relationship of the classes with the outcomes. Next, similar to quadratic discriminant analysis, we propose including a quadratic term in the logistic model for the LCs when the variances of the outcome are heteroskedastic in order to prevent parameter bias. And lastly we propose two alternative SE estimators (non-parametric bootstrap, jackknife), that yield better coverage rates than the currently used SE estimator proposed by Asparouhov and Muth{\'e}n (2014) . The proposed improvements are tested via a simulation study with good results, and applied to real data.
    },
    Author = {Z. Bakk and D. L. Oberski and J.K. Vermunt},
    Date-Modified = {2015-08-30 15:10:09 +0000},
    Doi = {10.1080/10705511.2015.1049698},
    Issue = {2},
    Journal = {Structural Equation Modeling: A Multidisciplinary Journal},
    Title = {Relating latent class membership to continuous distal outcomes: improving the {LTB} approach and a modified three-step implementation},
    Volume = {23},
    Year = {2016},
    Bdsk-Url-1 = {http://dx.doi.org/10.1080/10705511.2015.1049698}}

  • Oberski, D. L.. (2016). Beyond the number of classes: separating substantive from non-substantive dependence in latent class analysis. Advances in data analysis and classification, 10(2), 171-182. doi:10.1007/s11634-015-0211-0
    [BibTeX] [Abstract] [Download PDF] [Download data and code from the paper]
    Latent class analysis (LCA) for categorical data is a model-based clustering and classification technique applied in a wide range of fields including the social sciences, machine learning, psychiatry, public health, and epidemiology. Its central assumption is conditional independence of the indicators given the latent class, i.e. ``local independence''; violations can appear as model misfit, often leading LCA practitioners to increase the number of classes. However, when not all of the local dependence is of substantive scientific interest this leads to two options, that are both problematic: modeling uninterpretable classes, or retaining a lower number of substantive classes but incurring bias in the final results and classifications of interest due to remaining assumption violations. This paper suggests an alternative procedure, applicable in cases when the number of substantive classes is known in advance, or when substantive interest is otherwise well-defined. I suggest, in such cases, to model substantive local dependencies as additional discrete latent variables, while absorbing nuisance dependencies in additional parameters. An example application to the estimation of misclassification and turnover rates of the decision to vote in elections of 9510 Dutch residents demonstrates the advantages of this procedure relative to increasing the number of classes.
    @Article{oberski2016beyond,
    author = {Oberski, D. L.},
    title = {Beyond the number of classes: separating substantive from non-substantive dependence in latent class analysis},
    journal = {Advances in Data Analysis and Classification},
    year = {2016},
    volume = {10},
    number = {2},
    pages = {171-182},
    abstract = {
    Latent class analysis (LCA) for categorical data is a model-based clustering and classification technique applied in a wide range of fields including the social sciences, machine learning, psychiatry, public health, and epidemiology. Its central assumption is conditional independence of the indicators given the latent class, i.e. ``local independence''; violations can appear as model misfit, often leading LCA practitioners to increase the number of classes. However, when not all of the local dependence is of substantive scientific interest this leads to two options, that are both problematic: modeling uninterpretable classes, or retaining a lower number of substantive classes but incurring bias in the final results and classifications of interest due to remaining assumption violations.
    This paper suggests an alternative procedure, applicable in cases when the number of substantive classes is known in advance, or when substantive interest is otherwise well-defined. I suggest, in such cases, to model substantive local dependencies as additional discrete latent variables, while absorbing nuisance dependencies in additional parameters. An example application to the estimation of misclassification and turnover rates of the decision to vote in elections of 9510 Dutch residents demonstrates the advantages of this procedure relative to increasing the number of classes.
    },
    bdsk-url-1 = {http://daob.nl/wp-content/uploads/2015/03/beyond-the-number-of-classes.pdf},
    datapackage = {http://osf.io/mh278},
    date-added = {2014-10-16 13:54:15 +0000},
    date-modified = {2015-04-15 14:16:09 +0000},
    doi = {10.1007/s11634-015-0211-0},
    }

  • Nagelkerke, E., Oberski, D. L., & Vermunt, J. K.. (2016). Power and type i error of local fit statistics in multilevel latent class analysis. Structural equation modeling: a multidisciplinary journal. doi:10.1080/10705511.2016.1250639
    [BibTeX] [Abstract] [Download PDF]
    In the social and behavioral sciences, variables are often categorical and people nested in groups. Models for such data, such as multilevel logistic regression or the multilevel latent class model, should account for not only the categorical nature of the variables, but also the nested structure of the persons. To assess whether the model accomplishes this goal adequately, local fit measures for multilevel categorical data were recently introduced by Nagelkerke, Oberski, and Vermunt (2015). The "BVR-group" evaluates the variable-group fit, while the "BVR-pair" evaluates the person-person fit within groups. In this article, we evaluate the performance of these two measures for the multilevel latent class model (Vermunt, 2003). An extensive simulation study indicates that, whenever multilevel latent class modeling itself is viable, Type I error is controlled and power adequate for both fit statistics. Thus, the BVR-group and BVR-pair are useful measures to locate important sources of misfit in multilevel latent class analysis.
    @Article{nagelkerke2016power,
    author = {Nagelkerke, E. and Oberski, D.L. and Vermunt, J.K.},
    title = {Power and Type I Error of Local Fit Statistics in Multilevel Latent Class Analysis},
    journal = {Structural Equation Modeling: A Multidisciplinary Journal},
    year = {2016},
    abstract = {
    In the social and behavioral sciences, variables are often categorical and
    people nested in groups. Models for such data, such as multilevel logistic
    regression or the multilevel latent class model, should account for not only
    the categorical nature of the variables, but also the nested structure of the
    persons. To assess whether the model accomplishes this goal adequately,
    local fit measures for multilevel categorical data were recently introduced
    by Nagelkerke, Oberski, and Vermunt (2015). The "BVR-group" evaluates the variable-group fit, while the "BVR-pair" evaluates the person-person fit within groups. In this article, we evaluate the performance of
    these two measures for the multilevel latent class model (Vermunt, 2003).
    An extensive simulation study indicates that, whenever multilevel latent
    class modeling itself is viable, Type I error is controlled and power adequate for both fit statistics. Thus, the BVR-group and BVR-pair are useful measures to locate important sources of misfit in multilevel latent class analysis.
    },
    doi = {10.1080/10705511.2016.1250639},
    keywords = {bivariate residual, latent class analysis, local fit, multilevel},
    }

  • EFSA Panel on Animal Health Αnd Welfare (AHAW) HEALTHY-B working group. (2016). Assessing the health status of managed honeybee colonies (HEALTHY-B): a toolbox to facilitate harmonised data collection. EFSA journal, 14(10). doi:10.2903/j.efsa.2016.4578
    [BibTeX] [Download PDF]
    @Article{efsa2016,
    author = {{EFSA Panel on Animal Health Αnd Welfare (AHAW) HEALTHY-B working group}},
    title = {Assessing the health status of managed honeybee colonies ({HEALTHY-B}): a toolbox to facilitate harmonised data collection},
    journal = {{EFSA} Journal},
    year = {2016},
    volume = {14},
    number = {10},
    doi = {10.2903/j.efsa.2016.4578},
    url = {http://onlinelibrary.wiley.com/doi/10.2903/j.efsa.2016.4578/full},
    }

2015

  • Oberski, D. L., & Vermunt, J. K.. (2015). The relationship between CUB and loglinear models with latent variables. Electronic journal of applied statistical analysis, 08, 368-377. doi:10.1285/i20705948v8n3p368
    [BibTeX] [Abstract] [Download PDF] [Download data and code from the paper]
    The "combination of uniform and shifted binomial" (CUB) model is a distribution for ordinal variables that has received considerable recent attention and specialized development. This article notes that the CUB model is a special case of the well-known loglinear latent class model, an observation that is useful for two reasons. First, we show how it can be used to estimate the cub model in familiar standard software such as Mplus or Latent gold. Second, the mathematical equivalence of cub with this well-known model and its correspondingly long history allows well-known results to be applied straightforwardly, subsuming a wide range of specialized recent developments of cub and suggesting several possibly useful future ones. Thus, the observation that cub and its extensions are restricted loglinear latent class models should be useful to both applied practitioners and methodologists.
    @article{oberski2015cub,
    Abstract = {The "combination of uniform and shifted binomial" (CUB) model is a distribution for ordinal variables that has received considerable recent attention and specialized development. This article notes that the CUB model is a special case of the well-known loglinear latent class model, an observation that is useful for two reasons. First, we show how it can be used to estimate the cub model in familiar standard software such as Mplus or Latent gold. Second, the mathematical equivalence of cub with this well-known model and its correspondingly long history allows well-known results to be applied straightforwardly, subsuming a wide range of specialized recent developments of cub and suggesting several possibly useful future ones. Thus, the observation that cub and its extensions are restricted loglinear latent class models should be useful to both applied practitioners and methodologists.},
    Author = {Oberski, D. L. and Vermunt, J.K.},
    Datapackage = {http://daob.nl/wp-content/uploads/2015/03/Online-appendix.zip},
    Doi = {10.1285/i20705948v8n3p368},
    Issue = {03},
    Journal = {Electronic Journal of Applied Statistical Analysis},
    Pages = {368-377},
    Title = {The relationship between {CUB} and loglinear models with latent variables},
    Volume = {08},
    Year = {2015},
    Bdsk-Url-1 = {http://dx.doi.org/10.1285/i20705948v8n3p368}}

  • Meyers, C., Van Woerkom, M., De Reuver, R., Bakk, Z., & Oberski, D. L.. (2015). Enhancing psychological capital and personal growth initiative: developing strengths or deficiencies?. Journal of counseling psychology, 62. doi:10.1037/cou0000050
    [BibTeX] [Abstract]
    Personal growth initiative (PGI), defined as being proactive about one's personal development, is critical to graduate students' academic success. Prior research has shown that students' PGI can be enhanced through interventions that focus on stimulating developmental activities. Within this study, we aimed to investigate whether an intervention that stimulates development in the area of one's personal strengths (strengths intervention) has more beneficial effects on students' PGI than an intervention that stimulates development in the area of individual deficiencies (deficiency intervention). We conducted 2 longitudinal field experiments to investigate the effects of the 2 interventions on students' PGI (Experiment 1) and the potential mediating role of psychological capital (PsyCap) in this regard (Experiment 2). In Experiment 1, 105 (N = 105) university students participated in either a strengths intervention or a deficiency intervention. Results indicated that the strengths intervention increased the students' PGI in the short but not in the long term, whereas the deficiency intervention did not affect PGI. Ninety students (N = 90) participated in Experiment 2, in which we slightly refined both interventions by putting a stronger emphasis on the ongoing development of strengths (strengths intervention) or correction of deficiencies (deficiency intervention) by adding posttraining assignments. Results suggested that participating in both interventions led to increases in PGI over a 3-month period, but that these increases were bigger for the strengths intervention group. Furthermore, the relationship between the strengths intervention and PGI was mediated by hope as one component of PsyCap.
    @article{meyers2015enhancing,
    Abstract = {
    Personal growth initiative (PGI), defined as being proactive about one's personal development, is critical to graduate students' academic success. Prior research has shown that students' PGI can be enhanced through interventions that focus on stimulating developmental activities. Within this study, we aimed to investigate whether an intervention that stimulates development in the area of one's personal strengths (strengths intervention) has more beneficial effects on students' PGI than an intervention that stimulates development in the area of individual deficiencies (deficiency intervention). We conducted 2 longitudinal field experiments to investigate the effects of the 2 interventions on students' PGI (Experiment 1) and the potential mediating role of psychological capital (PsyCap) in this regard (Experiment 2). In Experiment 1, 105 (N = 105) university students participated in either a strengths intervention or a deficiency intervention. Results indicated that the strengths intervention increased the students' PGI in the short but not in the long term, whereas the deficiency intervention did not affect PGI. Ninety students (N = 90) participated in Experiment 2, in which we slightly refined both interventions by putting a stronger emphasis on the ongoing development of strengths (strengths intervention) or correction of deficiencies (deficiency intervention) by adding posttraining assignments. Results suggested that participating in both interventions led to increases in PGI over a 3-month period, but that these increases were bigger for the strengths intervention group. Furthermore, the relationship between the strengths intervention and PGI was mediated by hope as one component of PsyCap.
    },
    Author = {Meyers, C. and Van Woerkom, M. and De Reuver, R. and Bakk, Zs. and Oberski, D. L.},
    Doi = {10.1037/cou0000050},
    Issue = {1},
    Journal = {Journal of Counseling Psychology},
    Title = {Enhancing Psychological Capital and Personal Growth Initiative: Developing Strengths or Deficiencies?},
    Volume = {62},
    Year = {2015},
    Bdsk-Url-1 = {http://dx.doi.org/10.1037/cou0000050}}

  • Cieciuch, J., Davidov, E., Oberski, D. L., & Algesheimer, R.. (2015). Testing for measurement invariance by detecting local misspecification and an illustration across online and paper-and-pencil samples. European political science, 14(4), 521-538. doi:10.1057/eps.2015.64
    [BibTeX] [Abstract] [Download PDF]
    Public opinion researchers often need to evaluate whether groups can be compared, for example when comparing countries, constructing time series, or conducting mixedor multi-mode data collection. We review existing methods of doing so using structural equation models and reiterate arguments for their use in public opinion research. In addition, since public opinion questions are often categorical, we extend one recently introduced method of evaluating local model misspecifications to encompass categorical data. The methods are demonstrated by a novel comparison of data collection modes for measuring respondents' value priorities. This analysis shows that mixedor multi-mode mail and web data collection of respondents' value priorities is feasible, with the exception of the "societal security" value.
    @article{cieciuch2015testing,
    Abstract = {Public opinion researchers often need to evaluate whether groups can be compared, for example when comparing countries, constructing time series, or conducting mixedor multi-mode data collection. We review existing methods of doing so using structural equation models and reiterate arguments for their use in public opinion research. In addition, since public opinion questions are often categorical, we extend one recently introduced method of evaluating local model misspecifications to encompass categorical data. The methods are demonstrated by a novel comparison of data collection modes for measuring respondents' value priorities. This analysis shows that mixedor multi-mode mail and web data collection of respondents' value priorities is feasible, with the exception of the "societal security" value.},
    Author = {J. Cieciuch and E. Davidov and D. L. Oberski and R. Algesheimer},
    Date-Modified = {2015-11-24 08:50:23 +0000},
    Doi = {10.1057/eps.2015.64},
    Journal = {European Political Science},
    Number = {4},
    Pages = {521-538},
    Title = {Testing for measurement invariance by detecting local misspecification and an illustration across online and paper-and-pencil samples},
    Volume = {14},
    Year = {2015},
    Bdsk-Url-1 = {http://dx.doi.org/10.1057/eps.2015.64}}

  • Oberski, D. L., Vermunt, J. K., & Moors, G. B. D.. (2015). Evaluating measurement invariance in categorical data latent variable models with the EPC-interest. Political analysis, 23(4), 550-563. doi:10.1093/pan/mpv020
    [BibTeX] [Abstract] [Download PDF] [Download data and code from the paper]
    Many variables crucial to the social sciences are not directly observed but instead are latent and measured indirectly. When an external variable of interest affects this measurement, estimates of its relationship with the latent variable will then be biased. Such violations of "measurement invariance" may, for example, confound true differences across countries in postmaterialism with measurement differences. To deal with this problem, researchers commonly aim at "partial measurement invariance", i.e. to account for those differences that may be present and important. To evaluate this importance directly through sensitivity analysis, the "EPC-interest" was recently introduced for continuous data. However, latent variable models in the social sciences often use categorical data. The current paper therefore extends the EPC-interest to latent variable models for categorical data and demonstrates its use in example analyses of US Senate votes as well as respondent rankings of postmaterialism values in the World Values Study.
    @article{oberski2015EPC-interest-categorical,
    Abstract = {
    Many variables crucial to the social sciences are not directly observed but instead are latent and measured indirectly. When an external variable of interest affects this measurement, estimates of its relationship with the latent variable will then be biased. Such violations of "measurement invariance" may, for example, confound true differences across countries in postmaterialism with measurement differences. To deal with this problem, researchers commonly aim at "partial measurement invariance", i.e. to account for those differences that may be present and important. To evaluate this importance directly through sensitivity analysis, the "EPC-interest" was recently introduced for continuous data. However, latent variable models in the social sciences often use categorical data. The current paper therefore extends the EPC-interest to latent variable models for categorical data and demonstrates its use in example analyses of US Senate votes as well as respondent rankings of postmaterialism values in the World Values Study.
    },
    Author = {D. L. Oberski and J.K. Vermunt and G.B.D. Moors},
    Datapackage = {http://daob.nl/wp-content/uploads/2015/07/oberski-etal-replication-materials.tar.gz},
    Date-Modified = {2015-10-08 15:08:31 +0000},
    Doi = {10.1093/pan/mpv020},
    Journal = {Political Analysis},
    Number = {4},
    Pages = {550-563},
    Title = {Evaluating measurement invariance in categorical data latent variable models with the {EPC}-interest},
    Url = {http://pan.oxfordjournals.org/cgi/reprint/mpv020?ijkey=B2tbikjLW3ft0dp&keytype=ref},
    Volume = {23},
    Year = {2015},
    Bdsk-Url-1 = {http://dx.doi.org/10.1093/pan/mpv020}}

  • Nagelkerke, E., Oberski, D. L., & Vermunt, J. K.. (2015). Goodness-of-fit of multilevel latent class models for categorical data. Sociological methodology, 46(1). doi:10.1177/0081175015581379
    [BibTeX] [Abstract] [Download PDF]
    In the context of multilevel latent class models, the goodness-of-fit depends on multiple aspects, among which are two local independence assumptions. However, because of the lack of local fit statistics, the model and any issues relating to model fit can only be inspected jointly through global fit statistics. This hinders the search for model improvements, as it cannot be determined where misfit originates and which of the many model adjustments may improve its fit. Also, when relying solely on global fit statistics, assumption violations may become obscured, leading to wrong substantive results. In this paper, two local fit statistics are proposed to improve the understanding of the model, allow individual testing of the local independence assumptions, and inspect the fit of the higher level of the model. Through an application in which the local fit statistics group-variable residual and paired-case residual are used as guidance, it is shown that they pinpoint misfit, enhance the search for model improvements, provide substantive insight, and lead to a model with different substantive conclusions, which would likely not have been found when relying on global information criteria. Both residuals can be obtained in the user-friendly Latent GOLD 5.0 software package.
    @Article{nagelkerke2015goodness,
    author = {Nagelkerke, E. and Oberski, D. L. and Vermunt, J.K.},
    title = {Goodness-of-fit of Multilevel Latent Class Models for Categorical Data},
    journal = {Sociological Methodology},
    year = {2015},
    volume = {46},
    number = {1},
    abstract = {
    In the context of multilevel latent class models, the goodness-of-fit depends on multiple aspects, among which are two local independence assumptions. However, because of the lack of local fit statistics, the model and any issues relating to model fit can only be inspected jointly through global fit statistics. This hinders the search for model improvements, as it cannot be determined where misfit originates and which of the many model adjustments may improve its fit. Also, when relying solely on global fit statistics, assumption violations may become obscured, leading to wrong substantive results. In this paper, two local fit statistics are proposed to improve the understanding of the model, allow individual testing of the local independence assumptions, and inspect the fit of the higher level of the model. Through an application in which the local fit statistics group-variable residual and paired-case residual are used as guidance, it is shown that they pinpoint misfit, enhance the search for model improvements, provide substantive insight, and lead to a model with different substantive conclusions, which would likely not have been found when relying on global information criteria. Both residuals can be obtained in the user-friendly Latent GOLD 5.0 software package.
    },
    bdsk-url-1 = {http://dx.doi.org/10.1177/0081175015581379},
    doi = {10.1177/0081175015581379},
    }

  • Oberski, D. L., Hagenaars, J., & Saris, W. E.. (2015). The latent class multitrait-multimethod model. Psychological methods, 20, 422-443. doi:10.1037/a0039783
    [BibTeX] [Abstract] [Download PDF]
    A "latent class multitrait-multimethod" (MTMM) model is proposed to estimate random and systematic measurement error in categorical survey questions while making fewer assumptions than have been made so far in such evaluations, allowing for possible extreme response behavior and other nonmonotone effects. The method is a combination of the multitrait-multimethod research design of Campbell & Fiske (1959), the basic response model for survey questions of Saris & Andrews (1991),, and the latent class factor model of Vermunt & Magidson (2004, 227--230).. The latent class MTMM model thus combines an existing design, model, and method to allow for the estimation of the degree to and manner in which survey questions are affected by systematic measurement error. Starting from a general form of the response function for a survey question, we present the multitrait-multimethod (MTMM) experimental approach to identification of the response function's parameters. A ``trait-method biplot'' is introduced as a means of interpreting the estimates of systematic measurement error, while the quality of the questions can be evaluated by item information curves and the item information function, which we derive in the appendix. An experiment from the European Social Survey (ESS) is analyzed and the results are discussed, yielding valuable insights into the functioning of a set of example questions on the role of women in society in two countries.
    @article{oberski:WP:latent-class-mtmm,
    Abstract = {
    A "latent class multitrait-multimethod" (MTMM) model is proposed to
    estimate random and
    systematic measurement error in categorical survey questions while making fewer
    assumptions than have been made so far in such evaluations, allowing for
    possible extreme response behavior and other nonmonotone effects.
    The method is a combination of the multitrait-multimethod research design of
    Campbell & Fiske (1959), the basic response model for survey questions
    of Saris & Andrews (1991),, and the latent class factor model of
    Vermunt & Magidson (2004, 227--230).. The latent class MTMM model thus
    combines an existing design, model, and method to allow for the estimation
    of the degree to and manner in which survey questions are affected by systematic
    measurement error.
    Starting from a general form of the response function for a survey question,
    we present the multitrait-multimethod (MTMM) experimental approach to identification of
    the response function's parameters. A ``trait-method biplot'' is introduced
    as a means of interpreting the estimates of systematic measurement error, while
    the quality of the questions can be evaluated by item information curves
    and the item information function, which we derive in the appendix.
    An experiment from the European Social Survey (ESS) is analyzed and the
    results are discussed, yielding valuable insights into the
    functioning of a set of example questions on the role of
    women in society in two countries.
    },
    Author = {Oberski, D. L. and Hagenaars, J. and Saris, W.E.},
    Doi = {10.1037/a0039783},
    Issue = {4},
    Journal = {Psychological Methods},
    Pages = {422--443},
    Title = {The Latent Class Multitrait-Multimethod Model},
    Url = {http://daob.nl/wp-content/uploads/2013/03/Oberski-Hagenaars-Saris-latent-class-MTMM.pdf},
    Volume = {20},
    Year = {2015},
    Bdsk-Url-1 = {http://daob.nl/wp-content/uploads/2013/03/Oberski-Hagenaars-Saris-latent-class-MTMM.pdf}}

2014

  • Oberski, D. L.. (2014). Evaluating sensitivity of parameters of interest to measurement invariance in latent variable models. Political analysis, 22(1), 45-60. doi:10.1093/pan/mpt014
    [BibTeX] [Abstract] [Download PDF] [Download data and code from the paper]
    Latent variable models can only be compared across groups when these groups exhibit measurement equivalence or ``invariance'', since otherwise substantive differences may be confounded with measurement differences. This article suggests examining directly whether measurement differences present could confound substantive analyses, by examining the EPC-interest. The EPC-interest approximates the change in parameters of interest that can be expected when freeing cross-group invariance restrictions. Monte Carlo simulations suggest that the EPC-interest approximates these changes well. Three empirical applications show that the EPC-interest can help avoid two undesirable situations: first, it can prevent unnecessarily concluding that groups are incomparable, and second, it alerts the user when comparisons of interest may still be invalidated even when the invariance model appears to fit the data. R code and data for the examples discussed in this article are provided in the electronic appendix <a href="http://hdl.handle.net/1902.1/21816">http://hdl.handle.net/1902.1/21816</a>.
    @article{oberski2014EPCinterest,
    Abstract = {
    Latent variable models can only be compared across groups when these groups exhibit measurement equivalence or ``invariance'', since otherwise substantive differences may be confounded with measurement differences. This article suggests examining directly whether measurement differences present could confound substantive analyses, by examining the EPC-interest. The EPC-interest approximates the change in parameters of interest that can be expected when freeing cross-group invariance restrictions. Monte Carlo simulations suggest that the EPC-interest approximates these changes well. Three empirical applications show that the EPC-interest can help avoid two undesirable situations: first, it can prevent unnecessarily concluding that groups are incomparable, and second, it alerts the user when comparisons of interest may still be invalidated even when the invariance model appears to fit the data.
    R code and data for the examples discussed in this article are provided in the electronic appendix <a href="http://hdl.handle.net/1902.1/21816">http://hdl.handle.net/1902.1/21816</a>.
    },
    Author = {Oberski, D. L.},
    Datapackage = {http://hdl.handle.net/1902.1/21816},
    Date-Added = {2014-04-08 09:30:04 +0000},
    Date-Modified = {2014-04-08 09:30:28 +0000},
    Doi = {10.1093/pan/mpt014},
    Journal = {Political Analysis},
    Number = {1},
    Pages = {45-60},
    Title = {Evaluating Sensitivity of Parameters of Interest to Measurement Invariance in Latent Variable Models},
    Volume = {22},
    Year = {2014},
    Bdsk-Url-1 = {http://dx.doi.org/10.1093/pan/mpt014}}

  • Bakk, Z., Oberski, D. L., & Vermunt, J. K.. (2014). Relating latent class assignments to external variables: standard errors for correct inference. Political analysis, 22, 520-540. doi:10.1093/pan/mpu003
    [BibTeX] [Abstract] [Download PDF]
    Latent class analysis is used in the political science literature in both substantive applications and as a tool to estimate measurement error. Many studies in the social and political sciences relate estimated class assignments from a latent class model to external variables. Although common, such a "three-step" procedure effectively ignores classification error in the class assignments; Vermunt (2010, "Latent class modeling with covariates: Two improved three-step approaches," Political Analysis 18:450--69) showed that this leads to inconsistent parameter estimates and proposed a correction. Although this correction for bias is now implemented in standard software, inconsistency is not the only consequence of classification error. We demonstrate that the correction method introduces an additional source of variance in the estimates, so that standard errors and confidence intervals are overly optimistic when not taking this into account. We derive the asymptotic variance of the third-step estimates of interest, as well as several candidate-corrected sample estimators of the standard errors. These corrected standard error estimators are evaluated using a Monte Carlo study, and we provide practical advice to researchers as to which should be used so that valid inferences can be obtained when relating estimated class membership to external variables.
    @article{bakk:WP:variance,
    Abstract = {
    Latent class analysis is used in the political science literature in both substantive applications and as a tool to estimate measurement error. Many studies in the social and political sciences relate estimated class assignments from a latent class model to external variables. Although common, such a "three-step" procedure effectively ignores classification error in the class assignments; Vermunt (2010, "Latent class modeling with covariates: Two improved three-step approaches," Political Analysis 18:450--69) showed that this leads to inconsistent parameter estimates and proposed a correction. Although this correction for bias is now implemented in standard software, inconsistency is not the only consequence of classification error. We demonstrate that the correction method introduces an additional source of variance in the estimates, so that standard errors and confidence intervals are overly optimistic when not taking this into account. We derive the asymptotic variance of the third-step estimates of interest, as well as several candidate-corrected sample estimators of the standard errors. These corrected standard error estimators are evaluated using a Monte Carlo study, and we provide practical advice to researchers as to which should be used so that valid inferences can be obtained when relating estimated class membership to external variables.
    },
    Author = {Bakk, Z. and Oberski, D. L. and Vermunt, J.K.},
    Date-Added = {2013-06-30 11:35:25 +0000},
    Date-Modified = {2014-04-08 09:28:16 +0000},
    Doi = {10.1093/pan/mpu003},
    Journal = {Political Analysis},
    Pages = {520--540},
    Title = {Relating latent class assignments to external variables: standard errors for correct inference},
    Url = {http://daob.nl/wp-content/uploads/2013/06/bakk-oberski-vermunt-Relating-latent-class-assignments-to-external-variables-standard-errors-for-corrected-inference-2013.pdf},
    Volume = {22},
    Year = {2014},
    Bdsk-Url-1 = {http://daob.nl/wp-content/uploads/2013/06/bakk-oberski-vermunt-Relating-latent-class-assignments-to-external-variables-standard-errors-for-corrected-inference-2013.pdf},
    Bdsk-Url-2 = {http://dx.doi.org/10.1093/pan/mpu003}}

  • Oberski, D. L.. (2014). lavaan.survey: an R package for complex survey analysis of structural equation models. Journal of statistical software, 57(1), 1-27. doi:10.18637/jss.v057.i01
    [BibTeX] [Download PDF] [Download data and code from the paper]
    @article{oberski_lavaan,
    Author = {Oberski, D. L.},
    Datapackage = {http://www.jstatsoft.org/v57/i01},
    Date-Modified = {2013-07-29 06:54:51 +0000},
    Journal = {Journal of Statistical Software},
    Number = {1},
    Pages = {1--27},
    doi = {10.18637/jss.v057.i01},
    Title = {{lavaan.survey}: An {R} Package for Complex Survey Analysis of Structural Equation Models},
    Url = {http://www.jstatsoft.org/v57/i01/paper},
    Volume = {57},
    Year = {2014},
    Bdsk-Url-1 = {http://www.jstatsoft.org/v57/i01}}

2013

  • Oberski, D. L., & Vermunt, J. K.. (2013). A model-based approach to goodness-of-fit evaluation in item response theory. Measurement: interdisciplinary research & perspectives, 11, 117-122. doi:10.1080/15366367.2013.835195
    [BibTeX] [Download PDF]
    @article{Oberski:WP:model-based-GOF,
    Author = {Oberski, D. L. and Vermunt, J.K.},
    Date-Added = {2013-07-28 19:00:56 +0000},
    Date-Modified = {2013-09-18 01:42:24 +0000},
    Doi = {10.1080/15366367.2013.835195},
    Journal = {Measurement: Interdisciplinary Research & Perspectives},
    Pages = {117-122},
    Title = {A Model-Based Approach to Goodness-of-Fit Evaluation in Item Response Theory},
    Url = {http://daob.nl/wp-content/uploads/2013/07/Oberski-Vermunt-A-model-based-approach-to-goodness-of-fit-evaluation-in-item-response-theory-2013-07-28.pdf},
    Volume = {11},
    Year = {2013},
    Bdsk-Url-1 = {http://daob.nl/wp-content/uploads/2013/07/Oberski-Vermunt-A-model-based-approach-to-goodness-of-fit-evaluation-in-item-response-theory-2013-07-28.pdf},
    Bdsk-Url-2 = {http://dx.doi.org/10.1080/15366367.2013.835195}}

  • Oberski, D. L., {Van} Kollenburg, G. H., & Vermunt, J. K.. (2013). A Monte Carlo evaluation of three methods to detect local dependence in binary data latent class models. Advances in data analysis and classification, 7(3). doi:10.1007/s11634-013-0146-2
    [BibTeX] [Download PDF] [Download data and code from the paper]
    @article{oberski:WP:lca-bvr,
    Abtract = {Binary data latent class analysis is a form of model-based clustering applied in a wide range of fields. A central assumption of this model is that of conditional independence of responses given latent class membership, often referred to as the "local independence" assumption. The results of latent class analysis may be severely biased when this crucial assumption is violated; investigating the degree to which bivariate relationships between observed variables fit this hypothesis therefore provides vital information. This article evaluates three methods of doing so. The first is the commonly applied method of referring the so-called "bivariate residuals" to a chi-square distribution. We also introduce two alternative methods that are novel to the investigation of local dependence in latent class analysis: bootstrapping the bivariate residuals, and the asymptotic score test or "modification index". A Monte Carlo simulation indicates that the latter two methods perform adequately, while the first method does not perform as intended.},
    Author = {Oberski, D. L. and {Van} Kollenburg, G.H. and Vermunt, J.K.},
    Datapackage = {https://osf.io/cbrks/},
    Date-Added = {2013-01-06 15:07:24 +0000},
    Date-Modified = {2013-07-29 06:54:36 +0000},
    Doi = {10.1007/s11634-013-0146-2},
    Journal = {Advances in Data Analysis and Classification},
    Number = {3},
    Title = {A {Monte Carlo} evaluation of three methods to detect local dependence in binary data latent class models},
    Url = {http://daob.nl/wp-content/uploads/2013/03/Oberski-vanKollenburg-Vermunt-A-Monte-Carlo-evaluation-of-three-methods-to-detect-local-dependence-in-binary-data-latent-class-models.pdf},
    Volume = {7},
    Year = {2013},
    Bdsk-Url-1 = {http://daob.nl/wp-content/uploads/2013/03/Oberski-vanKollenburg-Vermunt-A-Monte-Carlo-evaluation-of-three-methods-to-detect-local-dependence-in-binary-data-latent-class-models.pdf},
    Bdsk-Url-2 = {http://dx.doi.org/10.1007/s11634-013-0146-2}}

  • Oberski, D. L., & Satorra, A.. (2013). Measurement error models with uncertainty about the error variance. Structural equation modeling, 20, 409-428. doi:10.1080/10705511.2013.797820
    [BibTeX] [Download PDF] [Download data and code from the paper]
    @article{oberski2013measurement,
    Author = {Oberski, D. L. and Satorra, A.},
    Datapackage = {http://daob.nl/wp-content/uploads/2014/11/code_uncertainty_measerr_paper.zip},
    Doi = {10.1080/10705511.2013.797820},
    Issn = {1070-5511},
    Journal = {Structural Equation Modeling},
    Pages = {409-428},
    Publisher = {Taylor \& Francis},
    Title = {Measurement error models with uncertainty about the error variance},
    Url = {http://daob.nl/wp-content/uploads/2013/03/Oberski-Satorra-Measurement-error-models-with-uncertainty-about-the-error-variance-2013.pdf},
    Volume = {20},
    Year = {2013},
    Bdsk-Url-1 = {http://daob.nl/wp-content/uploads/2013/03/Oberski-Satorra-Measurement-error-models-with-uncertainty-about-the-error-variance-2013.pdf},
    Bdsk-Url-2 = {http://dx.doi.org/10.1080/10705511.2013.797820}}

2012

  • Gallego, A., & Oberski, D. L.. (2012). Personality and political participation: the mediation hypothesis. Political behavior, 34, 425-451. doi:10.1007/s11109-011-9168-7
    [BibTeX] [Download PDF]
    @article{gallego2012personality,
    Author = {Gallego, A. and Oberski, D. L.},
    Doi = {10.1007/s11109-011-9168-7},
    Gsid = {bZ_q2WE7XqwJ},
    Issue = {3},
    Journal = {Political Behavior},
    Pages = {425--451},
    Publisher = {Springer},
    Title = {Personality and Political Participation: The Mediation Hypothesis},
    Url = {http://daob.nl/wp-content/uploads/2013/03/Gallego-Oberski-Personality-and-Political-Participation-The-Mediation-Hypothesis-2011.pdf},
    Volume = {34},
    Year = {2012},
    Bdsk-Url-1 = {http://daob.nl/wp-content/uploads/2013/03/Gallego-Oberski-Personality-and-Political-Participation-The-Mediation-Hypothesis-2011.pdf},
    Bdsk-Url-2 = {http://dx.doi.org/10.1007/s11109-011-9168-7}}

2008

  • Coromina, L., Saris, W. E., & Oberski, D. L.. (2008). The quality of the measurement of interest in the political issues in the media in the ess. Ask research & methods.
    [BibTeX] [Download PDF]
    @article{coromina2008quality,
    Author = {L. Coromina and W.E. Saris and D. L. Oberski},
    Gsid = {13342000007653642616},
    Isbn = {9789033468292},
    Journal = {ASK Research \& Methods},
    Publisher = {Polish Academy of Sciences},
    Title = {The Quality of the Measurement of Interest in the Political Issues in the Media in the ESS},
    Url = {http://daob.nl/wp-content/uploads/2013/03/Coromina_Saris_Oberski_IPIMedia_2008.pdf},
    Year = {2008},
    Bdsk-Url-1 = {http://daob.nl/wp-content/uploads/2013/03/Coromina_Saris_Oberski_IPIMedia_2008.pdf}}