A Nondegenerate Penalized Likelihood Estimator for Variance Parameters in Multilevel Models
- 720 Downloads
Group-level variance estimates of zero often arise when fitting multilevel or hierarchical linear models, especially when the number of groups is small. For situations where zero variances are implausible a priori, we propose a maximum penalized likelihood approach to avoid such boundary estimates. This approach is equivalent to estimating variance parameters by their posterior mode, given a weakly informative prior distribution. By choosing the penalty from the log-gamma family with shape parameter greater than 1, we ensure that the estimated variance will be positive. We suggest a default log-gamma(2,λ) penalty with λ→0, which ensures that the maximum penalized likelihood estimate is approximately one standard error from zero when the maximum likelihood estimate is zero, thus remaining consistent with the data while being nondegenerate. We also show that the maximum penalized likelihood estimator with this default penalty is a good approximation to the posterior median obtained under a noninformative prior.
Our default method provides better estimates of model parameters and standard errors than the maximum likelihood or the restricted maximum likelihood estimators. The log-gamma family can also be used to convey substantive prior information. In either case—pure penalization or prior information—our recommended procedure gives nondegenerate estimates and in the limit coincides with maximum likelihood as the number of groups increases.
Key wordsBayes modal estimation hierarchical linear model mixed model multilevel model penalized likelihood variance estimation weakly informative prior
The research reported here was supported by the Institute of Education Sciences (grant R305D100017) and the National Science Foundation (SES-1023189), the Department of Energy (DE-SC0002099), and National Security Agency (H98230-10-1-0184).
- Bates, D., & Maechler, M. (2010). lme4: Linear mixed-effects models using S4 classes. R. package version 0.999375-37. Google Scholar
- Bell, W. (1999). Accounting for uncertainty about variances in small area estimation. In Bulletin of the International Statistical Institute, 52nd session, Helsinki. Google Scholar
- Box, G., & Cox, D. (1964). An analysis of transformations. Journal of the Royal Statistical Society. Series B, 26(2), 211–252. Google Scholar
- Browne, W., & Draper, D. (2006). A comparison of Bayesian and likelihood methods for fitting multilevel models. Bayesian Analysis, 1(3), 473–514. Google Scholar
- Crainiceanu, C., Ruppert, D., & Vogelsang, T. (2003). Some properties of likelihood ratio tests in linear mixed models (Technical report). Available at http://www.orie.cornell.edu/~davidr/papers.
- Dorie, V. (2013). Mixed methods for mixed models: Bayesian point estimation and classical uncertainty measures in multilevel models. PhD thesis, Columbia University. Google Scholar
- Dorie, V., Liu, J., & Gelman, A. (2013). Bridging between point estimation and Bayesian inference for generalized linear models (Technical report). Department of Statistics, Columbia University. Google Scholar
- Draper, D. (1995). Assessment and propagation of model uncertainty. Journal of the Royal Statistical Society. Series B, 57(1), 45–97. Google Scholar
- Gelman, A. (2006). Prior distributions for variance parameters in hierarchical models. Bayesian Analysis, 1(3), 515–533. Google Scholar
- Gelman, A., Carlin, J., Stern, H., & Rubin, D. (2004). Bayesian data analysis (2nd ed.). London: Chapman & Hall/CRC. Google Scholar
- Huber, P.J. (1967). The behavior of maximum likelihood estimation under nonstandard condition. In L.M. LeCam & J. Neyman (Eds.), Proceedings of the fifth Berkeley symposium on mathematical statistics and probability (Vol. 1, pp. 221–233). Berkeley: University of California Press. Google Scholar
- Morris, C. (2006). Mixed model prediction and small area estimation (with discussions). Test, 15(1), 72–76. Google Scholar
- Rabe-Hesketh, S., & Skrondal, A. (2012). Multilevel and longitudinal modeling using Stata (3rd ed.). College Station: Stata Press. Google Scholar
- Verbeke, G., & Molenberghs, G. (2000). Linear mixed models for longitudinal data. Berlin: Springer. Google Scholar
- Vermunt, J., & Magidson, J. (2005). Technical guide for Latent Gold 4.0: basic and advanced (Technical report). Statistical Innovations Inc., Belmont, Massachusetts. Google Scholar
- Weiss, R.E. (2005). Modeling longitudinal data. New York: Springer. Google Scholar
- Whaley, S., Sigman, M., Neumann, C.G., Bwibo, N.O., Guthrie, D., Weiss, R.E., Alber, S., & Murphy, S.P. (2003). Animal source foods improve dietary quality, micronutrient status, growth and cognitive function in Kenyan school children: background, study design and baseline findings. The Journal of Nutrition, 133(11), 3965–3971. Google Scholar