1 Stata SEM LightStone 4 SEM 4.. Alan C. Acock, Discovering Structural Equation Modeling Using Stata, Revised Edition, Stata Press 3.

Transcription

1 1 Stata SEM LightStone 4 SEM 4.. Alan C. Acock, Discovering Structural Equation Modeling Using Stata, Revised Edition, Stata Press 3.

2 2 4, Depress Conservative. 3., 3,. SES66 Alien67 Alien71, Alien67 Alien71,. Wheaton et al. (1977),,. Wheaton,. Wheaton et al. (1977) Alan C. Acock, Discovering Structural Equation Modeling Using Stata, Revised Edition, Stata Press 3.

3 3 4 SES66,. educ66 occstat66. Alien67 Alien71. anonima, pwless(powerlessness) SES sem sm2.dta..use sem sm2,clear. (,, ).. Stata Summary Statistics Data(SSD),.,.

4 4.1. SES 4 Endogenous variables Measurement: Latent: Exogenous variables Latent: educ66 occstat66 anomia67 pwless67 anomia71 pwless71 Alien67 Alien71 SES66 Fitting target model: ( ) Structural equation model Number of obs = 932 Estimation method = ml Log likelihood = ( 1) [anomia67]alien67 = 1 ( 2) [anomia71]alien71 = 1 ( 3) [educ66]ses66 = 1 OIM Standardized Coef. Std. Err. z P> z [95% Conf. Interval] Structural Alien67 <- SES Alien71 <- Alien SES Measurement educ66 <- SES _cons occstat66 <- SES _cons anomia67 <- Alien _cons pwless67 <- Alien _cons anomia71 <- Alien _cons pwless71 <- Alien _cons var(e.educ66) var(e.occstat66) var(e.anomia67) var(e.pwless67) var(e.anomia71) var(e.pwless71) var(e.alien67) var(e.alien71) var(ses66) 1... LR test of model vs. saturated: chi2(6) = 71.62, Prob > chi2 =

5 5 4.,., 3 1. ( 1) [anomia67]alien67 = 1 ( 2) [anomia71]alien71 = 1 ( 3) [educ66]ses66 = , 0.85,. Structural ( ), Alien67 Alien71, β = estat eqgof

6 4.1. SES 6 Equation-level goodness of fit Variance depvars fitted predicted residual R-squared mc mc2 observed educ occstat anomia pwless anomia pwless latent Alien Alien overall mc = correlation between depvar and its prediction mc2 = mc^2 is the Bentler-Raykov squared multiple correlation coefficient Alien %, Alien %. SEM,. Alien67 Alien71 β = 0.66,. SEM, anonima( ) (Alien),... estat gof,stats(all)

7 7 4 Fit statistic Value Description Likelihood ratio chi2_ms(6) model vs. saturated p > chi chi2_bs(15) baseline vs. saturated p > chi Population error RMSEA Root mean squared error of approximation 90% CI, lower bound upper bound pclose Probability RMSEA <= 0.05 Information criteria AIC Akaike s information criterion BIC Bayesian information criterion Baseline comparison CFI Comparative fit index TLI Tucker-Lewis index Size of residuals SRMR Standardized root mean squared residual CD Coefficient of determination chi2 ms(6),.,. RMSE ,. CFI ,. MI.

8 4.1. SES 8. estat mindices Modification indices Standard MI df P>MI EPC EPC Measurement educ66 <- anomia67 <- pwless67 <- anomia71 <- pwless71 <- anomia pwless educ anomia pwless educ anomia pwless anomia pwless anomia pwless cov(e.educ66,e.anomia67) cov(e.educ66,e.pwless67) cov(e.anomia67,e.anomia71) cov(e.anomia67,e.pwless71) cov(e.pwless67,e.anomia71) cov(e.pwless67,e.pwless71) EPC = expected parameter change MI,, anomia71 anomia67,. anomia powerless,,,,. anomia powerless,.,

9 9 4, SES66 Alien71 SES66 Alien67, Alien71 Alien67 Alien71 ( )

10 estat gof,stats(all), χ 2 (6) = 71.62, p = χ 2 (4) = 4.77, p = 0.31 RMSE=0.108 RMSE=0.014 CFI=0.97 CFI=1.000 MI.. estat teffects,nodirect ( ) Structural Alien67 SES66 0 (no path) Alien71 Alien67 0 (no path) SES Indirect effects, SES66 Alien SEM,., Stata SEM. Stata SSD(Summary statistics data)., SSD sem, gsem.

11 x1,x2,x3. 3, 5, , , clear all 2...ssd init x1 x2 x ssd set obs ssd set cov \ \ ssd set cor. 5...ssd set means ssd status 7...ssd list 8., replace..ssd set means ,replace 9.,. 10. SEM,.

12 SSD,. SSD SEM. 1. sem vce(sbentler) Satorra-Bentler, Satorra- Bentler χ sem vce(robust). 3. sem vce(cluster clustvar). 4. svy: sem. 5. sem vce(bootstrap) vce(jackknife). 6. sem vce(opg). 7., [fw=varname]. 8. if in. 9. method(mlmv) method(adf)...use sem sm2,clear.ssd list,,.

13 GSEM. 2 GSEM. GSEM,,,... webuse gsem 1fmm,clear. sum Variable Obs Mean Std. Dev. Min Max x x x x s x1 x4 4 / 0 1. x1 x3 100 x4 725 s4 4 x1 x4, X.. 2 Stata.

14 4.3. GSEM 14 SEM. GSEM SEM.

15 15 4,.., GSEM. Fitting fixed-effects model: Iteration 0: log likelihood = Iteration 1: log likelihood = Iteration 2: log likelihood = ( ) Generalized structural equation model Number of obs = 123 Response Family Link Response Family Link Response Family Link Response Family Link : x1 : Bernoulli : probit : x2 : Bernoulli : probit : x3 : Bernoulli : probit : x4 : Bernoulli : probit Log likelihood = ( 1) [x1]x = 1 Coef. Std. Err. z P> z [95% Conf. Interval] x1 <- x2 <- x3 <- x4 <- X 1 (constrained) _cons X _cons X _cons X _cons var(x) x1 x4. x4 s4.,.

16 4.3. GSEM 16.

17 17 4 Fitting fixed-effects model: ( ) Generalized structural equation model Number of obs = 123 Response Family Link Response Family Link Response Family Link Response Family Link : x1 : Bernoulli : probit : x2 : Bernoulli : probit : x3 : Bernoulli : probit : s4 : Gaussian : identity Log likelihood = ( 1) [x1]x = 1 Coef. Std. Err. z P> z [95% Conf. Interval] x1 <- x2 <- x3 <- s4 <- X 1 (constrained) _cons X _cons X _cons X _cons var(x) var(e.s4) SEM GSEM,,. Stata 15 [SEM] example 30g.,.