Abstract:
|
In this paper, we propose two semiparametric single index models for spatially-temporally correlated data. One model has the nonparametric function separable from spatially correlated random effects and time effects. We call this model semiparametric spatio-temporal separable single index model (SSTS-SIM), while the other does not separate the nonparametric function and spatially correlated random effects but separates the time effects, we call it semiparametric spatio-temporal nonseparable single index model (SSTN-SIM). Two algorithms based on Markov Chain Expectation Maximization algorithm are introduced to estimate the models parameters, spatial effects and times effects. The proposed models are applied to the mortality data set of six major cities in South Korea. The data covers the period from January, 2000 to December, 2007. It is found that Busan city has the highest mortality and Seoul and Daejeon have the lowest mortality. SSTS-SIM enforces the unknown mortality functions of all cities to have the same shape but SSTN-SIM is more flexible. In terms of estimation, SSTNSIM is better than SSTSSIM. In terms of prediction, in case we have enough data, SSTN-SIM is better.
|
ASA Meetings Department
732 North Washington Street, Alexandria, VA 22314
(703) 684-1221 • meetings@amstat.org
Copyright © American Statistical Association.