利用空间统计方法建立疾病风险模型
发布时间：2018-12-17 04:33
【摘要】： 疾病地图经常应用于疾病地理分布的研究,构成疾病地图的数据通常是计数型数据。几十年来,统计工作者对疾病风险的空间差异模型进行了深入细致的研究,无论在理论上还是应用上,都取得了许多优秀成果。 由于相邻的区域之间,疾病相对风险比较相似,通常使用随机效应模型的方法,通过向相邻的区域借力来估计区域相对风险,这样可以使估计值更为稳定。为进一步研究空间疾病风险的差异程度,我们提出了新的疾病风险结构模型。 论文的结构安排如下: 首先,介绍了空间统计的概况,回顾了疾病地图在空间流行病学上的发展历史,并结合论文的内容,给出了一些相关的基本概念及方法。 其次,我们提出了一个构建相关结构模型的新方法,将离散的区域疾病风险连续化,把区域总体的风险视为一个连续的风险曲面,这样我们不仅可以估计每个区域个体的相对风险,还可以估计区域总体的疾病风险。同时,出于方便计算的考虑,论文对区域相对风险的分布函数进行了近似分析。 再次,我们结合随机效应方法及回归模型的有关应用,建立了非空间和空间的疾病风险模型,分别给出了Possion-lognormal模型和联合模型,并针对先验分布函数的选择进行比较分析。 最后,为验证模型准确性和稳定性,我们借助Matlab软件对提出的模型进行模拟检验。在模拟实验中,使用了MCMC方法进行相关的随机抽样。
[Abstract]:Disease maps are often applied to the study of geographical distribution of diseases. In recent decades, statisticians have made a thorough and detailed study on the spatial difference model of disease risk, and many excellent results have been obtained both in theory and in application. Because the relative risk of disease is similar between adjacent regions, the method of stochastic effect model is usually used to estimate the relative risk of the region by borrowing from adjacent regions, which can make the estimated value more stable. In order to further study the difference of spatial disease risk, we propose a new disease risk structure model. The structure of the paper is as follows: firstly, the general situation of spatial statistics is introduced, and the development history of disease map in spatial epidemiology is reviewed. Combined with the contents of the paper, some basic concepts and methods are given. Secondly, we propose a new method to construct the relevant structural model, which makes the discrete regional disease risk continuous, and treats the regional overall risk as a continuous risk surface. In this way, we can estimate not only the relative risk of individuals in each region, but also the overall disease risk of the region. At the same time, for the convenience of calculation, the distribution function of regional relative risk is analyzed approximately. Thirdly, combining the application of stochastic effect method and regression model, we establish non-spatial and spatial disease risk models, give Possion-lognormal model and joint model, and compare and analyze the choice of prior distribution function. Finally, in order to verify the accuracy and stability of the model, we use Matlab software to test the proposed model. In the simulation experiment, the MCMC method is used to carry out the related random sampling.
【学位授予单位】：燕山大学
【学位级别】：硕士
【学位授予年份】：2008
【分类号】：R311
本文编号：2383661
[Abstract]:Disease maps are often applied to the study of geographical distribution of diseases. In recent decades, statisticians have made a thorough and detailed study on the spatial difference model of disease risk, and many excellent results have been obtained both in theory and in application. Because the relative risk of disease is similar between adjacent regions, the method of stochastic effect model is usually used to estimate the relative risk of the region by borrowing from adjacent regions, which can make the estimated value more stable. In order to further study the difference of spatial disease risk, we propose a new disease risk structure model. The structure of the paper is as follows: firstly, the general situation of spatial statistics is introduced, and the development history of disease map in spatial epidemiology is reviewed. Combined with the contents of the paper, some basic concepts and methods are given. Secondly, we propose a new method to construct the relevant structural model, which makes the discrete regional disease risk continuous, and treats the regional overall risk as a continuous risk surface. In this way, we can estimate not only the relative risk of individuals in each region, but also the overall disease risk of the region. At the same time, for the convenience of calculation, the distribution function of regional relative risk is analyzed approximately. Thirdly, combining the application of stochastic effect method and regression model, we establish non-spatial and spatial disease risk models, give Possion-lognormal model and joint model, and compare and analyze the choice of prior distribution function. Finally, in order to verify the accuracy and stability of the model, we use Matlab software to test the proposed model. In the simulation experiment, the MCMC method is used to carry out the related random sampling.
【学位授予单位】：燕山大学
【学位级别】：硕士
【学位授予年份】：2008
【分类号】：R311
【参考文献】
相关期刊论文 前2条
1 肖根如;程朋根;潘海燕;陈斐;;基于空间统计分析与GIS研究江西省县域经济[J];东华理工学院学报;2006年04期
2 鲁凤;徐建华;;中国区域经济差异的空间统计分析[J];华东师范大学学报(自然科学版);2007年02期
，本文编号：2383661
本文链接：https://www.wllwen.com/yixuelunwen/shiyanyixue/2383661.html
最近更新
教材专著
