|
Abstract:
|
Many studies have reported associations between ambient temperature and daily deaths in cities. The regional impacts of climate change will vary widely depending on the vulnerability of the population. The nonlinear relations (U, J, or V shaped) have been observed with increased mortality at both high and low temperatures. The effect of temperature and a change point in association with increased mortality has also been studied. However, these changes points and the number of change points vary by regions. The relationship between temperature and mortality are highly dependent on unknown change points, the methods for simultaneously identifying the relationship and detecting change points varying by regions are quite limited. Therefore, in this paper, we develop a spatially adapted multi-change point detection method under the Bayesian hierarchical framework. Unlike traditional approach which assumed fixed change points, we treat change points as random variables and select them via variable selection procedure. Since the relationship between temperature and mortality depends on random change points, our nonparametric function can be varied by region.
|
