王娇,李长平,胡良平.复杂抽样调查设计二值资料一水平多重Logistic回归分析[J].四川精神卫生杂志,2019,32(5):385-389.
复杂抽样调查设计二值资料一水平多重Logistic回归分析
One-level multiple Logistic regression analysis of the dichotomous choice data collected from the complex sampling survey design
投稿时间:2019-09-27  
DOI:10.11886/j.issn.1007-3256.2019.05.001
中文关键词:  复杂抽样  二值资料  Logistic回归分析  抽样权重  派生变量
英文关键词:Complex sampling  Binary data  Logistic regression analysis  Sampling weights  Derived variable
基金项目:国家高技术研究发展计划课题资助(2015AA020102)
作者单位邮编
王娇 天津医科大学公共卫生学院卫生统计学教研室 300070
李长平 天津医科大学公共卫生学院卫生统计学教研室世界中医药学会联合会临床科研统计学专业委员会 100029
胡良平 世界中医药学会联合会临床科研统计学专业委员会军事科学院研究生院 100850
摘要点击次数:
全文下载次数:
中文摘要:
      本文目的是介绍复杂抽样调查设计二值资料多重logistic回归分析方法。通过一个实例,利用八种不同的分析策略(不考虑抽样设计和抽样权重、考虑抽样设计不考虑抽样权重、不考虑抽样设计考虑抽样权重、同时考虑抽样设计和抽样权重以及分别不考虑与考虑派生变量)对数据进行建模。对所得结果进行比较得出如下结论 在对复杂抽样设计资料进行统计分析的过程中,同时考虑抽样设计和抽样权重可以得到符合数据内部变量间依赖关系真实情况的结论。此外,本研究还介绍了采用SAS软件中SURVEYLOGISTIC过程对复杂抽样调查数据进行多重Llogistic回归分析的详细步骤。
英文摘要:
      The purpose of this paper was to introduce the method of multiple logistic regression analysis for binary data of complex sampling survey design. Eight different analysis strategies (regardless of sampling design and sampling weights; considering sampling design without considering sampling weights; without considering sampling design but considering sampling weights, and considering both sampling design and sampling weights, and then considering the derived variables under the four situations mentioned before, respectively) were used to model and analyze the survey data. By comparing the results, the following conclusions were drawn: in the process of statistical analysis of complex sampling design data, the conclusions obtained by considering sampling design and sampling weights were more in line with the real situation of the dependence between internal variables of data. In addition, this study also introduced the detailed steps of using SURVEYLOGISTIC procedure in SAS software to carry out multiple logistic regression analysis of complex sampling survey data.
      附件
查看全文  HTML  查看/发表评论  下载PDF阅读器
关闭
手机扫一扫