典型相关分析的过程和程序.docx

典型相关分析的过程和程序

第一篇：典型相关分析的过程和程序典型相关程序：proccancorrdata=work.xingvprefix=uwprefix=v;varx1-x2;withy1-y2;run;结果：（典型相关系数及检验）CanonicalCorrelationAnalysisAdjustedApproximateSquaredCanonicalCanonicalStandardCanonicalCorrelationCorrelationErrorCorrelation0.7885080.7746980.0772110.6217450.053740.0.2035350.002888此结果可知，第一对典型变量之间的典型相关系数为0.788508，它比0.053740大，说明正确。EigenvaluesofInv(E)*H=CanRsq/(1-CanRsq)EigenvalueDifferenceProportionCumulative1.64371.64080.99820.99820.00290.00181.0000下面为用似然比法检验典型相关系数与0的差别是否显著。TestofH0:ThecanonicalcorrelationsinthecurrentrowandallthatfollowarezeroLikelihoodApproximateRatioFValueNumDFDenDFPr>F0.377162886.600.00030.997112040.060.8031由于0.00030.05，所以说明第二对典型相关系数不显著。所以只取一对典型变量。TheCANCORRProcedure（原始的典型系数）CanonicalCorrelationAnalysisRawCanonicalCoefficientsfortheVARVariablesu1u2X1X10.0565661954-0.139971093X2X20.07073683130.1869496027RawCanonicalCoefficientsfortheWITHVariablesv1v2y1y10.0502425983-0.176147939y2y20.08022239880.2620835635用原指标来线性表达第一对典型变量的系数：U1=0.0565661954x1+0.0707368313x2V1=0.0502425983y1+0.0802223988y2CanonicalCorrelationAnalysis（标准化的典型系数）StandardizedCanonicalCoefficientsfortheVARVariablesu1u2X1X10.5522-1.3664X2X20.52151.3784StandardizedCanonicalCoefficientsfortheWITHVariablesv1v2y1y10.5044-1.7686y2y20.53831.7586用标准化指标来线性表达第一对典型变量的系数，即：U1=0.5522x1+0.5215x2V1=0.5044y1+0.5383y2第一典型变量在x1和x2上的系数几乎一样大。同理。TheCANCORRProcedure（典型结构）CanonicalStructureCorrelationsBetweentheVARVariablesandTheirCanonicalVariablesu1u2X1X10.9353-0.3539X2X20.92720.3747X1在典型变量u1上的系数和它跟典型变量u1的相关系数符号相同，都是正的。所以x1对u1有正的影响。同理x2对u1也有正的影响。CorrelationsBetweentheWITHVariablesandTheirCanonicalVariablesv1v2y1y10.9562-0.2927y2y20.96160.2743CorrelationsBetweentheVARVariablesandtheCanonicalVariablesoftheWITHVariablesv1v2X1X10.7375-0.0190X2X20.73110.0201CorrelationsBetweentheWITHVariablesandtheCanonicalVariablesoftheVARVariablesu1u2y1y10.7540-0.0157y2y20.75830.0147第二篇：典型相关分析典型相关分析在ＳPＳＳ中可以有两种方法来拟合典型相关分析，第一种是采用Manova过程来拟合，第二种是采用专门提供的宏程序来拟合，第二种方法在使用上非常简单，而输出的结果又非常详细，因此这里只对他进行介绍。该程序名为Canonical