Help Interpreting Results Using "nocons" Command

[Ray Hartman]

Hello,

I am a little confused about the "nocons" command. As I understand, it will drop the constant term. I am using ibn. since I want keep all the variables in the model, bu the requires using "noconstant". When I run this command:

global country "lngdppc_lag ibn.polity_demo2"    
global individual "male age age_sq married divorced religious income college_edu" 
global opt ""       
asdoc meqrlogit fight $individual $country $opt, noconstant || ccodecow:, cov(un), nested save(H3 - The Effect of Level of Democraticness on Willingness to Fight) replace setstars(***@.001, **@.01, *@.05) tzok dec(3) cnames(Model 1) title(H3 - The Effect of Level of Democraticness on Willingness to Fight)

I get the following results which do not have a _cons, but the user-created "asdoc" command reports this term: lns1_1_1:_cons. However, I do not understand what this means, and if it needs to be reported or not.

Mixed-effects logistic regression               Number of obs     =    310,418
Group variable: ccodecow                        Number of groups  =         92

                                                Obs per group:
                                                              min =        371
                                                              avg =    3,374.1
                                                              max =     13,637

Integration points =   7                        Wald chi2(12)     =    9503.87
Log likelihood = -162742.26                     Prob > chi2       =     0.0000

------------------------------------------------------------------------------
       fight |      Coef.   Std. Err.      z    P>|z|     [95% Conf. Interval]
-------------+----------------------------------------------------------------
        male |   .6906793   .0088812    77.77   0.000     .6732725    .7080861
         age |   .0137731   .0013196    10.44   0.000     .0111867    .0163595
      age_sq |  -.0002313   .0000141   -16.36   0.000     -.000259   -.0002036
     married |   .2968113   .0887262     3.35   0.001     .1229111    .4707115
    divorced |    .175669   .0887919     1.98   0.048     .0016401    .3496979
   religious |  -.6418332   .0166147   -38.63   0.000    -.6743975    -.609269
      income |  -.0023552   .0016195    -1.45   0.146    -.0055294    .0008189
 college_edu |  -.0719106   .0100945    -7.12   0.000    -.0916955   -.0521258
 lngdppc_lag |  -.6262017   .0173671   -36.06   0.000    -.6602406   -.5921628
             |
polity_demo2 |
          1  |   6.699032   .2104093    31.84   0.000     6.286637    7.111426
          2  |   6.590388   .2128285    30.97   0.000     6.173252    7.007524
          3  |   6.451505   .2145495    30.07   0.000     6.030996    6.872014
------------------------------------------------------------------------------

------------------------------------------------------------------------------
  Random-effects Parameters  |   Estimate   Std. Err.     [95% Conf. Interval]
-----------------------------+------------------------------------------------
ccodecow: Identity           |
                  var(_cons) |    1.02656   .1550357      .7635409    1.380183
------------------------------------------------------------------------------
LR test vs. logistic model: chibar2(01) = 25271.80    Prob >= chibar2 = 0.0000

On a final note, I have heard it is important to include a constant term in the results table, but if I am using "nocons", what should I do? Also, does it matter if the constant term is not significant although the predictors and the model's overall fit are significant?

Any thoughts and/or suggestions are appreciated.

Cheers,

Ray
Stata 16MP

[Nick Cox]

noconstant is an option, not a command. I am here pinging Attaullah Shah who is the author of asdoc (from SSC), as your question seems mostly about that command.