I am trying to examine a cross-lagged model with twin data. Without taking into account the correlatedness between twins, the model I am fitting is the following:
sem (predictor_1 -> outcome_1, ) (predictor_1 -> Outcome_2, ) (Predictor_2 -> Outcome_2, ) if (METhourperday < 35 & METhourperday_V3 < 35), method(mlmv) nocapslatent
Endogenous variables
Observed: outcome_1 Outcome_2
Exogenous variables
Observed: predictor_1 Predictor_2
Fitting saturated model:
Iteration 0: log likelihood = -2262.2407
Iteration 1: log likelihood = -2262.1998
Iteration 2: log likelihood = -2262.1998
Fitting baseline model:
Iteration 0: log likelihood = -2677.6786
Iteration 1: log likelihood = -2677.6701
Iteration 2: log likelihood = -2677.6701
Fitting target model:
Iteration 0: log likelihood = -2265.3014
Iteration 1: log likelihood = -2265.3014
Structural equation model Number of obs = 803
Estimation method: mlmv
Log likelihood = -2265.3014
--------------------------------------------------------------------------------------
| OIM
| Coefficient std. err. z P>|z| [95% conf. interval]
---------------------+----------------------------------------------------------------
Structural |
outcome_1 |
predictor_1 | .7741317 .0222355 34.82 0.000 .730551 .8177125
_cons | -.1627265 .0174839 -9.31 0.000 -.1969942 -.1284587
-------------------+----------------------------------------------------------------
Outcome_2 |
predictor_1 | .0274648 .025986 1.06 0.291 -.0234669 .0783965
Predictor_2 | .3268096 .0333928 9.79 0.000 .2613609 .3922583
_cons | .2381121 .0267395 8.90 0.000 .1857036 .2905207
---------------------+----------------------------------------------------------------
var(e.outcome_1)| .1579486 .0080135 .1429981 .1744621
var(e.Outcome_2)| .2183358 .0108964 .1979905 .2407717
--------------------------------------------------------------------------------------
LR test of model vs. saturated: chi2(2) = 6.20 Prob > chi2 = 0.0450
. estat gof, stats(rmsea)
----------------------------------------------------------------------------
Fit statistic | Value Description
---------------------+------------------------------------------------------
Population error |
RMSEA | 0.051 Root mean squared error of approximation
90% CI, lower bound | 0.007
upper bound | 0.099
pclose | 0.399 Probability RMSEA <= 0.05
----------------------------------------------------------------------------
. estat gof, stats(ic)
----------------------------------------------------------------------------
Fit statistic | Value Description
---------------------+------------------------------------------------------
Information criteria |
AIC | 4544.603 Akaike's information criterion
BIC | 4577.421 Bayesian information criterion
----------------------------------------------------------------------------
. estat gof, stats(indices)
----------------------------------------------------------------------------
Fit statistic | Value Description
---------------------+------------------------------------------------------
Baseline comparison |
CFI | 0.995 Comparative fit index
TLI | 0.987 Tucker–Lewis index
----------------------------------------------------------------------------
When I try to take the twin data structure into account, I do the following:
svyset, psu(family)
svy linearized: sem (predictor_1 -> outcome_1, ) (predictor_1 -> Outcome_2, ) (Predictor_2 -> Outcome_2, ) if (METhourperday < 35 & METhourperday_V3 < 35), method(mlmv) nocapslatent
The model runs okay, but I cannot get model fitting indices (RMSEA, TLI, CFI, AIC, BIC). Is there a way to take into account the twin data structure AND get these model fitting indices (RMSEA, TLI, CFI, AIC, BIC)?
I appreciate all help.
0 Response to Correlated data and structural equation modelling
Post a Comment