공간상관성을 고려한 회귀계수의 베이지안 군집모형을이용한 국내 당뇨병 유병률 자료 분석

초록

In spatial regression modeling, it is commonly assumed that spatial random components are considered to explain the spatial dependency structures and regression coefficient is constant over the entire spatial domain. However, the regression coefficient may have spatial dependency structures and be different depending on the sub-regions. Recently, Lawson et al. (2014) proposed Bayesian discrete clustering methods of spatially dependent regression coefficients and applied them to cancer survival dataset. Bayesian hierarchical approach was utilized to explain the complicated spatial dependent structures. In this paper, we first analyze the diabetes prevalence data for the entire 252 administrative districts of South Korea in 2014 year using spatially-dependent regression coefficient clustering models. We evaluate the performance of the proposed spatial models with the non-spatial model.

키워드