Summary
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.
Outcomes reported
Referenced by Nature Communications British biodiversity scenarios as citation 92; likely supports topic area: methods / modelling / statistics. Topics: methods / modelling / statistics Evidence type: Modelling / projection Source report: Nature Communications British biodiversity scenarios Ref#: Nature Communications British biodiversity scenarios #92 Original: Chung, Y., Rabe-Hesketh, S., Dorie, V., Gelman, A. & Liu, J. A nondegenerate penalized likelihood estimator for variance parameters in multilevel models. Psychometrika 78, 685-709 (2013).
Dig deeper with Pulse AI.
Pulse AI has read the whole catalogue. Ask about this record, its theme, or how the findings apply to UK farming and policy — every answer cites the underlying studies.