Application Of Generalized Linear Mixed Model In Death Data Analysis

Generalized linear mixed models(GLMM’s) are an organic bond of generalized linear models(GLM’s) and linear mixed models(LMM’s). GLMM inherits the advantage of GLM and LMM and is suitable for fitting correlated discrete data. It is suitable to use GLMM in death data study, because the death data subjects to a binomial distribution and has correlation among different are as and time. This paper studied the death data in China by GLMM. With the different set of random factors, we get 7 models available, and the results are as follows: man’s death rate is larger than woman’s; after the age of 10,the older, the greater the death rate; before the age of 10, the younger, the greater the death rate; the three lowest death rate provinces are Shanghai, Beijing and Hainan; the three highest death rate provinces are Tibet, Yunnan and Qinghai; the death rate of 2000 is greater than 2010’s; the death rate in urban is lower than town, town is lower than rural; in the Northwest and Southwest of China, the death rate has a big gap between urban and rural areas; in the Northeast and North of China, the death rate, especially in urban and town, is higher than any other areas; in the South and East of China, the death rate is lower than any other areas; in the Central of China, the death rate, regardless of urban, town or rural, is always in an average level. The above results are useful for the government to make poverty alleviation and health care policy. At the same time, it provides a proof for insurance actuarial in different areas and time.