|
Abstract:
|
In this talk, we propose a semiparametric multi-state frailty survival model to analyze tooth-level life course data in caries research. The model specifically accommodates non-trivial design complications including intra-oral clustering, interval censoring, multiplicity of caries states, and does not impose a parametric restriction on the baseline intensities. We establish a connection between this model and the classical mixed-effects model to conduct estimation and inference using a penalized likelihood. More importantly, the connection with the mixed effects model allows us to view the penalty parameters as variance components which are simultaneously estimated with other model parameters. Numerical experiments demonstrate that the proposed methods perform very well in finite samples with small-to-moderate sizes. The practical utility of the model is illustrated using life course data from a unique longitudinal study of dental caries in low-income urban African-American children.
|
