第八章因子-新讲述.ppt

现代 实用 越野 保留四个因子？ 再生相关阵残差明显下降 八、确认性因子分析Confirmatory Factor Analysis Loosely Speaking … Allows the user to specify prior notions about the structure of the factor model Allows the user to test hypotheses about model parameters and to assess fit Prior Notions? About which variables load on which factors. About how correlated you think the underlying factors are. Recall, in the exploratory model the underlying factors are necessarily modeled as uncorrelated. Testing and Fit? With some prior notions fixed and maximum likelihood estimation employed, standard errors of parameter estimates are available. These are “asymptotic” – or valid with large samples. Likewise, formal fit statistics are available (typically chi-squared statistics) that allow model comparisons Exploratory Two-Factor Model 待估参数：?ij, ?i, i=1,…5, j=1,2 Specific weighting pattern hypothesized. Number of parameters to be estimated reduced Typical Confirmatory Model with 待估参数：?11, ?21, ?31, ?42, ?52 ,?12, ?i, i=1,…5 Schematically X1 X2 X3 X4 X5 h1 h2 h3 h4 h5 f1 f2 Wij =1 if xi and xj are hypothesized to load on the same factor Wij=cov(f1,f2) if xi and xj are hypothesized to load on different factors Covariance Model 确认性因子分析一般模型 Asymptotic Standard Error(标准误的一致估计量 For the comfirmatory factor analysis, we are able to obtain asymptotic standard error of the parameter estimates. This enables us to conduct statistical tests of the parameter values Goodness of Fit The statistics to test the goodness-of-fit of the model Goodness-of-fit index(GFI) Better greater than o.95 GFI adjusted for degrees of freedom (AGFI) Better greater than o.90 ?2 with p value H0:the proposed model fits as well as a perfect model P0.2 Measure Reliability测量信度 tells us whether a particular variable Xi does a good job of measuring the true underlying factor that it purports to measure Rule of thumb: ?i≥0.7 X1 Xi Xk h1 hi h3 f1 … … ?1 ?i ?k Reliability of the composite index tells us whether the measurement model is good enough to measure