分类数据中高维列联表可压缩性分析.pdfVIP

摘要 5、给出了两种分别基于互信息和信息熵对列联表变量重要性的排序方法， 发现从列联表可压缩性的角度，基于互信息的变量重要性排序方法更加准确。而 从变量含有的不确定信息多少的角度，基于信息熵的变量重要性排序方法更加准 确。 研究的成果对分类数据分析方法的研究深入发展做出新的贡献，对高维列联 表的可压缩性方法提供了一些重要可实现的途径。 关键词：列联表压缩；辛普森悖论；交互作用；互信息；信息熵 Ⅱ Abstract Abstract data isan toolfordata ofnominaldata Categoricalanalysisimportant analysis and ordereddata．In data table ofdataisa categoricalanalysis，contingencyanalysis commonandintuitivemethod．For researchers cases example，medicalcategorizeby and toestablish workers students age gender contingencytable；educationclassify by and toestablish age，genderfamilybackground contingencytable；economic researchersthesuccessorfailureofthe and classify enterprisesbyindustry，area initialinvestmenttoestablish researchers cases contingencytable；market categorize to andthe of toestablish accordingage，gendertendencycommodityconsumption contingencytable，etc． Thetraditional data methodis forthe categoricalanalysis mainly independence testof table．Asthe modelis forwardand contingency log-linear used， put widely data methodisoftenusedto the categoricalanalysis analyzehighwaycontingency theliteratureindomesticandoverseaslackofdetailed ofthe table，but analysishigh table．Becauseofthe ofthe