Pemodelan Kematian yang Diakibatkan oleh Pembunuhan dan Bunuh Diri di Lithuania pada Tahun 2003-2004

Abstract

Many daily problems can be assumed as stochastic processes. Therefore stochastic processes are very important subjects. One of them is modeling the geographical variation of mortality risk. This paper considers numerical issues of the empirical Bayesian approach model applied to the low rate estimation. The condition for nonsingularity of Bayesian estimation is given and the convenient iterative algorithm for the estimation is described. The clustering algorithm is also developed. It uses the property of Poisson-Gaussian model to treat probabilities of events in populations being the same, if the variance of probabilities is small. If the population size is not taken into account, then the estimation of relative risk (RR), which is only based on a few cases to produce a map, is not good. Empirical Bayes estimation has been proven to have smaller squared error than that of RR estimator. Integrating and maximizing likelihood functions is done by using a mathematical software. This paper refers mainly to the paper of Sakalauskas (2010) entitled On the Empirical Bayesian Approach for the Poisson-Gaussian model.