《SPSS-logistic回归分析.ppt》由会员分享，可在线阅读，更多相关《SPSS-logistic回归分析.ppt（42页珍藏版）》请在金锄头文库上搜索。
1、logistic 回归回归海南医学院公共卫生学院海南医学院公共卫生学院卫生统计学教研室卫生统计学教研室赵婵娟赵婵娟Evaluation only.Created with Aspose.Slides for .NET 3.5 Client Profile 5.2.0.0.Copyright 2004-2011 Aspose Pty Ltd.内容内容l基本概念基本概念l基本步骤基本步骤l基本操作基本操作l基本结果解释基本结果解释Evaluation only.Created with Aspose.Slides for .NET 3.5 Client Profile 5.2.0.0.Copyri
2、ght 2004-2011 Aspose Pty Ltd.数据数据分析分析思路思路 l单因素统计分析单因素统计分析1.计量资料的比较，计量资料的比较，t检验，方差分析或秩和检验检验，方差分析或秩和检验2.两个变量的相关分析，两个变量的相关分析，Pearson相关或相关或spearman相关相关3.分类资料的比较，分类资料的比较，Pearson卡方检验卡方检验4.分类资料的相关，分类资料的相关，OR值及其值及其CI或列联系数或列联系数l多因素统计分析多因素统计分析1.因变量为计量资料，多重现性回归因变量为计量资料，多重现性回归2.因变量为分类变量，因变量为分类变量，logistic回归回归Eva
3、luation only.Created with Aspose.Slides for .NET 3.5 Client Profile 5.2.0.0.Copyright 2004-2011 Aspose Pty Ltd.Logistic回归回归模型模型按研究设计分类：按研究设计分类：1.非配对设计：非条件非配对设计：非条件logistic回归模型回归模型2.配对病例对照：条件配对病例对照：条件logistic回归模型回归模型按反应变量分类：按反应变量分类：1.二分类二分类logistic回归模型（常用）回归模型（常用）2.多分类无序多分类无序logistic回归模型（常用）回归模型（常用）3
4、.多分类有序多分类有序logistic回归模型（常用）回归模型（常用）Evaluation only.Created with Aspose.Slides for .NET 3.5 Client Profile 5.2.0.0.Copyright 2004-2011 Aspose Pty Ltd.基础知识基础知识l通过下例复习相关概念通过下例复习相关概念l如：研究患某疾病与饮酒的关联性如：研究患某疾病与饮酒的关联性l患病率患病率 P1=? P2=?Evaluation only.Created with Aspose.Slides for .NET 3.5 Client Profile 5.2
5、.0.0.Copyright 2004-2011 Aspose Pty Ltd.基础知识基础知识Evaluation only.Created with Aspose.Slides for .NET 3.5 Client Profile 5.2.0.0.Copyright 2004-2011 Aspose Pty Ltd.Evaluation only.Created with Aspose.Slides for .NET 3.5 Client Profile 5.2.0.0.Copyright 2004-2011 Aspose Pty Ltd.Evaluation only.Created
6、with Aspose.Slides for .NET 3.5 Client Profile 5.2.0.0.Copyright 2004-2011 Aspose Pty Ltd.二分类二分类logistic回归模型回归模型Evaluation only.Created with Aspose.Slides for .NET 3.5 Client Profile 5.2.0.0.Copyright 2004-2011 Aspose Pty Ltd.Evaluation only.Created with Aspose.Slides for .NET 3.5 Client Profile 5.2
7、.0.0.Copyright 2004-2011 Aspose Pty Ltd.Evaluation only.Created with Aspose.Slides for .NET 3.5 Client Profile 5.2.0.0.Copyright 2004-2011 Aspose Pty Ltd.回归系数的意义回归系数的意义l多因素多因素logistic回归分析时，对回归系数回归分析时，对回归系数的解释都是指在其他所有自变量固定的的解释都是指在其他所有自变量固定的情况下的优势比。情况下的优势比。l存在因素间交互作用时，存在因素间交互作用时，logistic回归系回归系数的解释变得更为
8、复杂，应特别小心。数的解释变得更为复杂，应特别小心。Evaluation only.Created with Aspose.Slides for .NET 3.5 Client Profile 5.2.0.0.Copyright 2004-2011 Aspose Pty Ltd.适用条件适用条件l因变量为二分类变量或某事件的发生率因变量为二分类变量或某事件的发生率l自变量与自变量与Logit（P）之间为线性关系）之间为线性关系l残差合计为残差合计为0，且服从二项分布，且服从二项分布l各观测间相互独立各观测间相互独立l参数估计方法：最大似然法参数估计方法：最大似然法Evaluation only
9、.Created with Aspose.Slides for .NET 3.5 Client Profile 5.2.0.0.Copyright 2004-2011 Aspose Pty Ltd.例例1l研究急性心肌梗塞（研究急性心肌梗塞（AMI）患病与饮酒的关）患病与饮酒的关系，采用横断面调查。系，采用横断面调查。Evaluation only.Created with Aspose.Slides for .NET 3.5 Client Profile 5.2.0.0.Copyright 2004-2011 Aspose Pty Ltd.SPSS基本操作基本操作Evaluation onl
10、y.Created with Aspose.Slides for .NET 3.5 Client Profile 5.2.0.0.Copyright 2004-2011 Aspose Pty Ltd.Evaluation only.Created with Aspose.Slides for .NET 3.5 Client Profile 5.2.0.0.Copyright 2004-2011 Aspose Pty Ltd.Evaluation only.Created with Aspose.Slides for .NET 3.5 Client Profile 5.2.0.0.Copyrig
11、ht 2004-2011 Aspose Pty Ltd.Evaluation only.Created with Aspose.Slides for .NET 3.5 Client Profile 5.2.0.0.Copyright 2004-2011 Aspose Pty Ltd.Evaluation only.Created with Aspose.Slides for .NET 3.5 Client Profile 5.2.0.0.Copyright 2004-2011 Aspose Pty Ltd.哑变量设置哑变量设置l为了便于解释，对二分类变量按为了便于解释，对二分类变量按0、1编码
12、编码l如果对二项分类变量按如果对二项分类变量按+1、-1编码，结果？编码，结果？l分类变量必须转化。如地区对血压的影响。分类变量必须转化。如地区对血压的影响。l等级资料，当等级之间量度不一时必须转化。等级资料，当等级之间量度不一时必须转化。l连续资料不宜直接进入方程时，转化为等级连续资料不宜直接进入方程时，转化为等级资料或分类资料。资料或分类资料。Evaluation only.Created with Aspose.Slides for .NET 3.5 Client Profile 5.2.0.0.Copyright 2004-2011 Aspose Pty Ltd.SPSS哑变量设置哑变
13、量设置Evaluation only.Created with Aspose.Slides for .NET 3.5 Client Profile 5.2.0.0.Copyright 2004-2011 Aspose Pty Ltd.Evaluation only.Created with Aspose.Slides for .NET 3.5 Client Profile 5.2.0.0.Copyright 2004-2011 Aspose Pty Ltd.Evaluation only.Created with Aspose.Slides for .NET 3.5 Client Profil
14、e 5.2.0.0.Copyright 2004-2011 Aspose Pty Ltd.Evaluation only.Created with Aspose.Slides for .NET 3.5 Client Profile 5.2.0.0.Copyright 2004-2011 Aspose Pty Ltd.Evaluation only.Created with Aspose.Slides for .NET 3.5 Client Profile 5.2.0.0.Copyright 2004-2011 Aspose Pty Ltd.矫正混杂作用矫正混杂作用l例例2：上例没有考虑吸烟情况
15、，故将吸烟作为：上例没有考虑吸烟情况，故将吸烟作为分层加入，资料如下：操作同例分层加入，资料如下：操作同例1Evaluation only.Created with Aspose.Slides for .NET 3.5 Client Profile 5.2.0.0.Copyright 2004-2011 Aspose Pty Ltd.逐步回归分析逐步回归分析Evaluation only.Created with Aspose.Slides for .NET 3.5 Client Profile 5.2.0.0.Copyright 2004-2011 Aspose Pty Ltd.逐步回归逐步
16、回归Evaluation only.Created with Aspose.Slides for .NET 3.5 Client Profile 5.2.0.0.Copyright 2004-2011 Aspose Pty Ltd.逐步逐步logistic回归回归Evaluation only.Created with Aspose.Slides for .NET 3.5 Client Profile 5.2.0.0.Copyright 2004-2011 Aspose Pty Ltd.Evaluation only.Created with Aspose.Slides for .NET 3.
17、5 Client Profile 5.2.0.0.Copyright 2004-2011 Aspose Pty Ltd.Evaluation only.Created with Aspose.Slides for .NET 3.5 Client Profile 5.2.0.0.Copyright 2004-2011 Aspose Pty Ltd.Evaluation only.Created with Aspose.Slides for .NET 3.5 Client Profile 5.2.0.0.Copyright 2004-2011 Aspose Pty Ltd.Evaluation o
18、nly.Created with Aspose.Slides for .NET 3.5 Client Profile 5.2.0.0.Copyright 2004-2011 Aspose Pty Ltd.Evaluation only.Created with Aspose.Slides for .NET 3.5 Client Profile 5.2.0.0.Copyright 2004-2011 Aspose Pty Ltd.条件条件logistic回归回归Evaluation only.Created with Aspose.Slides for .NET 3.5 Client Profi
19、le 5.2.0.0.Copyright 2004-2011 Aspose Pty Ltd.Evaluation only.Created with Aspose.Slides for .NET 3.5 Client Profile 5.2.0.0.Copyright 2004-2011 Aspose Pty Ltd.条件条件logistic回归回归Evaluation only.Created with Aspose.Slides for .NET 3.5 Client Profile 5.2.0.0.Copyright 2004-2011 Aspose Pty Ltd.logistic回归
20、模型小结回归模型小结Evaluation only.Created with Aspose.Slides for .NET 3.5 Client Profile 5.2.0.0.Copyright 2004-2011 Aspose Pty Ltd.logistic回归模型小结回归模型小结Evaluation only.Created with Aspose.Slides for .NET 3.5 Client Profile 5.2.0.0.Copyright 2004-2011 Aspose Pty Ltd.logistic回归模型小结回归模型小结Evaluation only.Create
21、d with Aspose.Slides for .NET 3.5 Client Profile 5.2.0.0.Copyright 2004-2011 Aspose Pty Ltd.小结小结Evaluation only.Created with Aspose.Slides for .NET 3.5 Client Profile 5.2.0.0.Copyright 2004-2011 Aspose Pty Ltd.谢谢大家！谢谢大家！Evaluation only.Created with Aspose.Slides for .NET 3.5 Client Profile 5.2.0.0.Copyright 2004-2011 Aspose Pty Ltd.
