1、.課 時(shí) 授 課 計(jì) 劃課次序號(hào)： 22一、課二、課題：實(shí)驗(yàn)九 典型相關(guān)分析型：上機(jī)實(shí)驗(yàn)三、目的要求：1.掌握典型相關(guān)分析的理論與方法、模型的建立與顯著性檢驗(yàn)；2.掌握利用典型相關(guān)分析的SAS過程解決有關(guān)實(shí)際問題.四、教學(xué)重點(diǎn)：典型相關(guān)分析的SAS過程.教學(xué)難點(diǎn)：相關(guān)分析的理論與方法、模型的建立與顯著性檢驗(yàn).五、教學(xué)方法及手段：傳統(tǒng)教學(xué)與上機(jī)實(shí)驗(yàn)相結(jié)合六、參考資料：使用統(tǒng)計(jì)方法與SAS多元統(tǒng)計(jì)分析(二版)，何曉群編，中國(guó)人民大學(xué)出版社，2008；應(yīng)用回歸分析(二版)，何曉群編，中國(guó)人民大學(xué)出版社，2007；統(tǒng)計(jì)建模與R七、作業(yè)：4.9 4.10八、授課記錄：班次-.九、授課效果分析：實(shí)驗(yàn)九 典

2、型相關(guān)分析(Canonical Correlation Analysis)一、實(shí)驗(yàn)?zāi)康暮鸵蠖?、?shí)驗(yàn)內(nèi)容1.典型相關(guān)分析的SAS過程PROC CANCORR過程PROC CANCORR ;VAR variables;WITH variables;RUN;F（1）proccancorr語句的： -. noprintshort simple vname為 vprefix為 wname為 wprefix為（2）VAR（3）WITHvariables與2. 典型相關(guān)分析步驟X( , , ) , Y ( , , )X X Y YXY ，TT12p12qx( , , ) ,x xy ( , , )y yx

3、yTT12p12qx ( , , ) ,x xy ( , , )xy yy，nTTi1i2ipi1i2iq-.x xipxx x xx (x ,x ,x ) (,)*1*i2*T11i22piips11ss22ppy yy y y yy (,)i ,n*11i22iqqTi111111RRR (X ,Y ) 1112 TTTRR2122樣本典型相關(guān)分析步驟：A R R R RB R R R R*1 1（*11 22 0212p*A R R R RB R R R R*1 1和11 e ,e ,ef , f ,f和12p12p第k1f R1y*2TTe R x, VU*，211kkkk22( ,

4、, , ( , , , )x x x xy y y y*2*2*1p1qk,p，kU*,V*kkH(k) :0 H( ): 0k ,pk0k1kd 11/t(0)真 ( , )H kFF d d2kkdk2kk1/kkk(P F)( ( ,) p fpfP F d dkH0kkk2kkp若H (k)0-. ,pkpp kk 1k 13.實(shí)例分析例 X : X : X :123Y :Y :Y :123 46nX ( , , ) , Y ( , , )X X X Y Y Y TT123123 cards;-.-.;run; ： data corr;/*調(diào)用相關(guān)分析過程，打印樣本相關(guān)系數(shù)矩陣*/-.

5、 run; R11RR12R由(X ,X ,X ,Y,Y ,Y )T R1231232122SAS 系統(tǒng)10:24 Sunday, November 2, 2008 14The CANCORR ProcedureCorrelations Among the Original VariablesCorrelationsAmongtheVARVariablesR ）x11.00000.57050.8751x20.57051.00000.7808x30.87510.78081.0000 x1x2x3Correlations Among the WITH VariablesR ）y11.00000.6

6、7050.7850y20.67051.00000.9324y30.78500.93241.0000y1y2y3Correlations Between the VAR Variables and the WITH VariablesRy1y2y3x1x2x30.71360.37960.62560.84000.68090.81850.91430.59070.8695變量間高度相關(guān)。SAS 系統(tǒng)10:24 Sunday, November 2, 2008 15The CANCORR ProcedureCanonical Correlation AnalysisAdjustedCanonicalCo

7、rrelationApproximateStandardErrorSquaredCanonicalCorrelationCanonicalCorrelation典型相關(guān)系數(shù)校正的典型相關(guān)系數(shù) 近似的標(biāo)準(zhǔn)誤 典型相關(guān)系數(shù)平方k2123 =0.9278570.9223450.5398330.0207330.1019580.144965 =0.86091911 =0.5621812 =0.31604722 =0.165974.=0.027547233-. Eigenvalues of Inv(E)*HTest of H0: The canonical correlations in the2 )=

8、CanRsq/(1-CanRsq) 即current row and all that follow are zeroLikelihood Approximate2kkEigenvalue Difference Proportion CumulativePr FRatioF Value Num DF Den DF Fddp各對(duì)相關(guān)系相鄰兩特特征值占 特征值占方差 似然比值kkk2kk數(shù)特征值6.1901征值之差5.7280方差比例 比例累計(jì)值10.92660.06920.00420.92660.99581.00000.092504400.665111580.9724526817.984.641

9、.199497.5 F檢驗(yàn)p值.0001統(tǒng)計(jì)方法F值Wilks LambdaPillais Trace0.0925044017.9891.204513669.399126.0001Hotelling-Lawley TraceRoys Greatest Root6.680470816.1900536029.1886.66959.75542.0001 F1108.911299.05759.16960.91180.08250.00570.9118 0.000497780.9943 0.054711581.0000 0.59381993141.5840.9417.789 58.56.000129.85

10、360.684045026 FWilks Lambda9.0001Pillais Trace9.0001Hotelling-Lawley TraceRoys Greatest Root119.44881991108.91116421934.66726.0001.00013NOTE: F Statistic for Roys Greatest Root is an upper bound.-.SAS 系統(tǒng) 2008年08月01日 星期五 下午09時(shí)25分34秒3The CANCORR ProcedureCanonical Correlation AnalysisRaw Canonical Coe

11、fficients for the VAR VariablesV10.0066747730.0318235747-0.002099295V20.0352304459-0.0124793340.0125781113V3x1x2x3-0.054341051-0.0051960290.0592150232Raw Canonical Coefficients for the WITH VariablesW10.02861776220.00084264310.0060022012W2-0.0304073750.04568545860.0139185651W3-0.066162702-0.04865502

12、30.116965176y1y2y3SAS 系統(tǒng) 2008年08月01日 星期五 下午09時(shí)25分34秒4The CANCORR Procedure 標(biāo)準(zhǔn)化數(shù)據(jù)典型相關(guān)分析Canonical Correlation AnalysisStandardized Canonical Coefficients for the VAR VariablesV10.14890.9770-0.0520V20.7857-0.38310.3116V3-1.2120-0.15951.4671x1x2x3Standardized Canonical Coefficients for the WITH Variable

13、sW10.85750.01930.1454W2-0.9111W3-1.9825y1y2y31.04630.3371-1.11432.8332A R R R R *11 0.995440 0.95282020.63732322,312第一特征根貢獻(xiàn)率U x x -x *1123V 0.8575y 0.0193y 0.1454y *1123-.0.995440 1U 0.7857x -x x *2123V y 1.0463y y *2123U x x x*3123V 1.9825y 1.1143y 2.8332y*3312 0.952820 0.63732332SAS 系統(tǒng)2008年08月01日

14、星期五 下午09時(shí)25分34秒5The CANCORR ProcedureCanonical StructureCorrelations Between the VAR Variables and Their Canonical VariablesV10.33250.99330.3827V20.9248-0.10080.7530V3-0.18470.05660.5352x1x2x3Correlations Between the WITH Variables and Their Canonical VariablesW10.99700.57100.9222W2-0.0647W3-0.0435y1y2y30.81120.2738-0.12580.2731Correlations Between the VAR Variables and the Canonical Variables of the WITH VariablesW10.33100