Multivariate model

[Ismail Adam]

Dear All,

I have a problem with the multivariate model, especially the likelihood ratio

global y1list select1
global y2list select2
global y3list select3
global y4list select4
global xlist AGE EDUC Market1a BGRADE EXPREJ Formality duration_ BPOWER FCONT

mvreg $y1list $y2list $y3list $y4list = $xlist in 176/377

How can I do the likelihood ratio test?

Best Regards,
Ismail

[Joseph Coveney]

How can I do the likelihood ratio test?

Here's one way. As an example, the likelihood-ratio test for the coefficients for AGE equal to zero is shown.

// Step 1
sem (select1 select2 select3 select4 <- ///
    AGE EDUC Market1a BGRADE EXPREJ Formality duration_ BPOWER FCONT) ///
    in 176/377, ///
        covstructure(e._OEn, unstructured) nocapslatent nocnsreport ///
            nodescribe nofootnote nolog
estimates store Unconstrained

// Step 2
quietly sem (select1 select2 select3 select4 <- ///
    AGE@0 EDUC Market1a BGRADE EXPREJ Formality duration_ BPOWER FCONT) ///
    in 176/377, ///
        covstructure(e._OEn, unstructured) nocapslatent
estimates store AGE_Equals_Zero

// Step 3
lrtest Unconstrained AGE_Equals_Zero

[Ismail Adam]

that doesn't work with me

[Ismail Adam]

.

[Ismail Adam]

I got this error message:

convergence not achieved
convergence not achieved
r(430);

[Joseph Coveney]

I got this error message:

convergence not achieved

Simplify your model. You can start by removing the covstructure(e._OEn, unstructured) or changing it to covstructure(e._OEn, exchangeable).

Follow-up as needed by removing predictors one at a time.

You can diagnose which predictor is giving you the most trouble by limiting the iteration tally and taking a look at the regression table. Use the iterate() option to do that.