结构方程模型LISREL肖前国.ppt

(一)题项间的区分效度的检验1.让题1和题2进行自由估计：ObservedVariables:jk1jk7jk13jk19jk25jk2jk8jk14jk20jk26jk3jk9jk15jk21jk27m1m2m3RawDatafromfileE:\LEARNING\SEM\difference.psfSamplesize=776LatentVariables:w1w2Relationships:jk1=1*w1jk7jk13jk19jk25=w1Setthecovariancesofjk1andjk7free(让jk1和jk7自由估计协方差)PathDiagramAdmissibilitycheck=offEndofProblem2.保存运行结果记录下卡方值和自由度：289.16,df=43.不让让题1和题2进行自由估计修改程序：ObservedVariables:jk1jk7jk13jk19jk25jk2jk8jk14jk20jk26jk3jk9jk15jk21jk27m1m2m3RawDatafromfileE:\LEARNING\SEM\difference.psfSamplesize=776LatentVariables:w1w2Relationships:jk1=1*w1jk7jk13jk19jk25=w1!Setthecovariancesofjk1andjk7free（加一个“！”就能使这条语句不运行）PathDiagramAdmissibilitycheck=offEndofProblem4.保存程序并运行结果记录卡方值和自由度：记录卡方值和自由度：286.48，df=55.利用Excel表格统计函数计算卡方差异具体：利用统计函数函数计算卡方分布的右尾概率，即P值6.解释结果结果表明这两个题项差异不显著，即这两个题项的区分效度不好。但这不意味着这个量表的区分效度不好！以此类推，可计算各分量表其他两两题项间的区分效度。(二)变量层面的区分效度的检验1计算多个变量间的测量模型的卡方值：ObservedVariables:jk1jk7jk13jk19jk25jk2jk8jk14jk20jk26jk3jk9jk15jk21jk27m1m2m3RawDatafromfileE:\LEARNING\SEM\difference.psfSamplesize=776LatentVariables:w1w2w3Relationships:jk1=1*w1jk7jk13jk19jk25=w1jk2=1*w2jk8jk14jk20jk26=w2jk3=1*w3jk9jk15jk21jk27=w3PathDiagramAdmissibilitycheck=offEndofProblem2计算多个变量间的测量模型的卡方值：ObservedVariables:jk1jk7jk13jk19jk25jk2jk8jk14jk20jk26jk3jk9jk15jk21jk27m1m2m3RawDatafromfileE:\LEARNING\SEM\difference.psfSamplesize=776LatentVariables:w1w2w3Relationships:jk1=1*w1jk7jk13jk19jk25=w1jk2=1*w2jk8jk14jk20jk26=w2jk3=1*w3jk9jk15jk21jk27=w3setthecorrelationbetweenw1andw2tobe1.0PathDiagramAdmissibilitycheck=offEndofProblem协方差：观察多个连续变量间的关联性（线性关系），反映的是多个变量间的“平均联合偏差”的大小。协方差等于相关系数乘以标准差。即如果变量均标准化后，协方差即相关系数。*误差方差为负：可能样本太少；或者公因子太少或太多；模型有问题。*残差：测定值与按回归方程预测的值之差/残差的标准差即为标准化残差；标准化残差大于2.58/3表明存在异常值。修正指标：若将原