Hi,

I have a question about the chow test for comparing coefficients in mixed models.

On this Stata FAQ pages, it talks about the chow test:

"You can include the dummy variables in a regression of the full model and then use the test command on those dummies. You could also run each of the models and then write down the appropriate numbers and calculate the statistic by hand—you also have access to functions to get appropriate p-values."

I am running two growth-curve models by males and females separately using stata mixed procedure. I then use chow test to see if the coefficients for the two mixed models are significantly different by males and females.

Is chow test appropriate for testing if the coefficients for the two mixed models are significantly different by males or females?

In my code:

g2 is my gender variable with g2: 1=female; 0=male; ctage1 is an age variable; I am running a cubic growth curve model.

My code for mixed model:

Code:
mixed cesd3w c.ctage1##i.g2 c.ctage1#c.ctage1##i.g2  ///
                       c.ctage1#c.ctage1#c.ctage1##i.g2 ///
                       [pweight=w1wt_fmch3] || aid: ctage1, pweight(schwt1) ///
                       pwscale(size) nolog cov(un) mle variance
The result for the mixed model:

Code:
Mixed-effects regression                        Number of obs     =     62,976
Group variable: aid                             Number of groups  =     18,185

                                                Obs per group:
                                                              min =          1
                                                              avg =        3.5
                                                              max =          5

                                                Wald chi2(7)      =     378.24
Log pseudolikelihood =  -20658221               Prob > chi2       =     0.0000

                                                (Std. Err. adjusted for 18,185 clusters in aid)
-----------------------------------------------------------------------------------------------
                              |               Robust
                       cesd3w |      Coef.   Std. Err.      z    P>|z|     [95% Conf. Interval]
------------------------------+----------------------------------------------------------------
                       ctage1 |   .0537449   .0158186     3.40   0.001     .0227409    .0847488
                         1.g2 |   .6205963   .0732295     8.47   0.000     .4770691    .7641235
                              |
                  g2#c.ctage1 |
                           1  |   .0001431   .0232488     0.01   0.995    -.0454237      .04571
                              |
            c.ctage1#c.ctage1 |  -.0050771   .0013704    -3.70   0.000     -.007763   -.0023912
                              |
         g2#c.ctage1#c.ctage1 |
                           1  |  -.0034937   .0020319    -1.72   0.086    -.0074762    .0004888
                              |
   c.ctage1#c.ctage1#c.ctage1 |   .0001352   .0000342     3.95   0.000     .0000682    .0002023
                              |
g2#c.ctage1#c.ctage1#c.ctage1 |
                           1  |   .0001173   .0000511     2.30   0.022     .0000172    .0002173
                              |
                        _cons |   .9459049   .0474442    19.94   0.000      .852916    1.038894
-----------------------------------------------------------------------------------------------

------------------------------------------------------------------------------
                             |               Robust           
  Random-effects Parameters  |   Estimate   Std. Err.     [95% Conf. Interval]
-----------------------------+------------------------------------------------
aid: Unstructured            |
                 var(ctage1) |   .0031306   .0003701       .002483     .003947
                  var(_cons) |   1.315501   .0724852      1.180835    1.465525
           cov(ctage1,_cons) |  -.0402105   .0046329     -.0492908   -.0311303
-----------------------------+------------------------------------------------
               var(Residual) |   1.926895   .0369594        1.8558    2.000713
------------------------------------------------------------------------------
My chow test:

Code:
 contrast g2 g2#c.ctage1  g2#c.ctage1#c.ctage1 g2#c.ctage1#c.ctage1#c.ctage1
My results for the chow test:

Code:
Contrasts of marginal linear predictions

Margins      : asbalanced

-----------------------------------------------------------------
                              |         df        chi2     P>chi2
------------------------------+----------------------------------
cesd3w                        |
                           g2 |          1       71.82     0.0000
                              |
                  g2#c.ctage1 |          1        0.00     0.9951
                              |
         g2#c.ctage1#c.ctage1 |          1        2.96     0.0855
                              |
g2#c.ctage1#c.ctage1#c.ctage1 |          1        5.28     0.0216
-----------------------------------------------------------------


By looking at the results from contrasts, can I say:

(1) the cubic age is statistically different by gender (g2#c.ctage1#c.ctage1#c.ctage1) with P=0.0216?

(2) But the linear and quadratic age are not significantly different by gender (g2#c.ctage1 with p=0.9951; g2#c.ctage1#c.ctage1 with p=0.0856)?

Thank you for your advice!

Alice