Dear all,

I am running the below 2 models in Stata. They have different dependent variables, and some similar independent variables (certain independent variables were removed since they were statistically insignificant). Below please see the code and results.

I am interested in a variable that is in both models – dar. In the first model, the coefficient on dar is -.062(P<0.001) and the second model the value of the coefficient on dar is 0.058 (P=0.001). I am trying to evaluate whether dar “explains” the first dependent variable or the second dependent variable better. In the first model, values of the dependent variable are mostly between 0 and 100, but with some values above 100 and below 0. In the second model, the values of dependent range from 0-100.

Can I simply compare the magnitude of the coefficients on dar determine this? For example, since the magnitude is larger in the first model, can I conclude that dar “explains” dependent variable 1 better? Is there a better way to do this, for example, just have dar as the only independent variable and compare BIC?


Thanks!!!


Code:
. mixed dep1 dar alp mup zol mupbyzol || _all: R.bor || _all: R.mol || _all: R.pop, reml

Performing EM optimization:

Performing gradient-based optimization:

Iteration 0:   log restricted-likelihood = -37391.644  
Iteration 1:   log restricted-likelihood = -37391.644  

Computing standard errors:

Mixed-effects REML regression                   Number of obs     =      8,476
Group variable: _all                            Number of groups  =          1

                                                Obs per group:
                                                              min =      8,476
                                                              avg =    8,476.0
                                                              max =      8,476

                                                Wald chi2(5)      =     226.97
Log restricted-likelihood = -37391.644          Prob > chi2       =     0.0000

------------------------------------------------------------------------------
        dep1 |      Coef.   Std. Err.      z    P>|z|     [95% Conf. Interval]
-------------+----------------------------------------------------------------
         dar |  -.0624097   .0153719    -4.06   0.000     -.092538   -.0322814
         alp |   8.150469   1.526269     5.34   0.000     5.159037     11.1419
         mup |   8.152124   2.433408     3.35   0.001     3.382732    12.92152
         zol |   6.486658   .7586855     8.55   0.000     4.999661    7.973654
    mupbyzol |  -1.694464    .915423    -1.85   0.064     -3.48866    .0997322
       _cons |   81.06523   2.629083    30.83   0.000     75.91233    86.21814
------------------------------------------------------------------------------

------------------------------------------------------------------------------
  Random-effects Parameters  |   Estimate   Std. Err.     [95% Conf. Interval]
-----------------------------+------------------------------------------------
_all: Identity               |
                  var(R.bor) |   63.10153   13.75592      41.16052    96.73842
-----------------------------+------------------------------------------------
_all: Identity               |
                  var(R.mol) |   18.28294   7.468788      8.209574     40.7166
-----------------------------+------------------------------------------------
_all: Identity               |
                  var(R.pop) |   30.08839   6.280492       19.9859    45.29751
-----------------------------+------------------------------------------------
               var(Residual) |   381.8593   5.912967      370.4442    393.6262
------------------------------------------------------------------------------
LR test vs. linear model: chi2(3) = 1810.17               Prob > chi2 = 0.0000

Note: LR test is conservative and provided only for reference.

.
. mixed dep2 dar alp mup zol alpbymup || _all: R.bor || _all: R.mol || _all: R.pop, reml

Performing EM optimization:

Performing gradient-based optimization:

Iteration 0:   log restricted-likelihood = -38246.114  
Iteration 1:   log restricted-likelihood = -38246.114  

Computing standard errors:

Mixed-effects REML regression                   Number of obs     =      8,463
Group variable: _all                            Number of groups  =          1

                                                Obs per group:
                                                              min =      8,463
                                                              avg =    8,463.0
                                                              max =      8,463

                                                Wald chi2(5)      =     246.42
Log restricted-likelihood = -38246.114          Prob > chi2       =     0.0000

------------------------------------------------------------------------------
        dep2 |      Coef.   Std. Err.      z    P>|z|     [95% Conf. Interval]
-------------+----------------------------------------------------------------
         dar |   .0576079   .0172886     3.33   0.001     .0237229    .0914928
         alp |  -9.863186   2.068844    -4.77   0.000    -13.91805   -5.808327
         mup |  -9.944255   3.193489    -3.11   0.002    -16.20338   -3.685131
         zol |  -6.253207   .4719679   -13.25   0.000    -7.178247   -5.328167
    alpbymup |  -2.215597   1.016369    -2.18   0.029    -4.207643   -.2235513
       _cons |   49.90762   3.607263    13.84   0.000     42.83752    56.97773
------------------------------------------------------------------------------

------------------------------------------------------------------------------
  Random-effects Parameters  |   Estimate   Std. Err.     [95% Conf. Interval]
-----------------------------+------------------------------------------------
_all: Identity               |
                  var(R.bor) |   185.1551   38.18656        123.59    277.3884
-----------------------------+------------------------------------------------
_all: Identity               |
                  var(R.mol) |   34.39981   14.81115      14.79325    79.99239
-----------------------------+------------------------------------------------
_all: Identity               |
                  var(R.pop) |   49.99171   10.22253      33.48424    74.63724
-----------------------------+------------------------------------------------
               var(Residual) |   470.1817   7.287957      456.1124     484.685
------------------------------------------------------------------------------
LR test vs. linear model: chi2(3) = 2918.94               Prob > chi2 = 0.0000

Note: LR test is conservative and provided only for reference.