广义线性模型在Python中的实现  

Implementation of generalized linear model in Python

作　　者：罗晨曦 张清煜 李湘莹 李培政 王静 马露[1] Chen-Xi LUO;Qing-Yu ZHANG;Xiang-Ying LI;Pei-Zheng LI;Jing WANG;Lu MA(School of Public Health,Wuhan University,Wuhan 430071,China)

机构地区：[1]武汉大学公共卫生学院,武汉430071

出　　处：《数理医药学杂志》2023年第3期161-165,共5页Journal of Mathematical Medicine

基　　金：湖北省卫生健康委2021—2022年度科研项目(WJ2021F103)。

摘　　要：目的探索广义线性模型(generalized linear model,GLM)在Python软件中的实现方法,并比较其与其他常用统计软件在算法过程和结果方面的异同。方法分别利用Python软件statsmodles库中的GLM函数、Logit和Poisson函数,R软件GLM函数,SAS的PROC GENMOD过程步,对二项分布和泊松分布的数据集进行分析,比较三种软件的算法过程和分析结果。结果三种软件构建GLM的逻辑相似,但在代码实现和模型拟合方法等方面稍有区别,各软件的结果基本相同。结论Python软件可采用不同的算法构建广义线性模型,并且能提供与其他主流统计软件相同的统计分析结论。Objective To explore the implementation method of generalized linear model in Python software,and compare its process and results with other common statistical software.Method Using GLM function,Logit and Poisson function in statsmodles library of Python software,GLM function in R software and PROC GENMOD procedure step of SAS,the binomial distribution and Poisson distribution data sets were analyzed,and the process and analysis results of the three software algorithms were compared.Results The logic of the three kinds of software to construct the generalized linear model is similar,but the software is slightly different in code implementation,model fitting methods and other aspects,and the results of the software are basically the same.Conclusion Python softwares can use different algorithms to build generalized linear models,and can provide the same statistical analysis conclusions as other mainstream statistical softwares.

关 键 词：PYTHON SAS R 广义线性模型 统计分析 

分 类 号：O212.1[理学—概率论与数理统计] TP312.1[理学—数学]