An Analysis of Type 2 Diabetes of Korean Adults via Generalized Partially Linear Varying Coefficient Model

The generalized partially linear varying coefficient models are natural extensions of generalized linear models and can be used in various applications. Penalized likelihood method is one of popular nonparametric methods for estimating unknown smooth functions. In this paper, we propose a simple method to estimate varying coefficient functions in a lower-dimensional approximating function space and regression parameters in partially linear part simultaneously by using penalized likelihood method. The proposed methods are applied to Korean diabetes analysis by using Korean national health and nutrition examination survey (KNHANES) data to investigate how the prevalence of type 2 diabetes of Korean adults is associated with physiological risk factors and is also correlated with age. Fitting the generalized partially linear varying coefficient model to the data showed that the diabetes prevalence was positively associated with triglyceride and their association was correlated with age.