Goodness-of-fit measure, mixed-effects GLM, meglm

[Hans Koslowski]

Dear all,

I have a little question with respect to the output of my regression below. I ran a mixed-effects GLM regression on several bond characteristics using the meglm command. I conducted a Gaussian distribution and the identity link as you can see in the output below, which I assume to be equivalent with the correlated random effects model aka the Mundlak model. Can anybody confirm this?

Additionally, the meglm command displays no classic goodness-of-fit measure, such as the (adjusted) R-square. Does anybody know whether there is any comparable goodness of fit measure in this output? Or do you guys know a way to run goodness-of-fit tests for GLMs in Stata?

Thank you guys very much and all the best,
Hans

. meglm YieldtoMaturityMid Standards2 ExternalReview2 Reporting2 BidAsk Ticker_num Currency_num PaymentRank_num MaturityYears2 AmountIssued

Iteration 0:   log likelihood = -25559.669  
Iteration 1:   log likelihood = -25559.669  

Mixed-effects GLM                               Number of obs     =      9,411
Family:                Gaussian
Link:                  identity

                                                Wald chi2(8)      =          .
Log likelihood = -25559.669                     Prob > chi2       =          .
-------------------------------------------------------------------------------------------
       YieldtoMaturityMid |      Coef.   Std. Err.      z    P>|z|     [95% Conf. Interval]
--------------------------+----------------------------------------------------------------
               Standards2 |  -1.354097   .4394719    -3.08   0.002    -2.215446   -.4927475
          ExternalReview2 |   .2656997   .4194891     0.63   0.526    -.5564838    1.087883
               Reporting2 |   .9717874   .2680178     3.63   0.000     .4464823    1.497093
                   BidAsk |   -.479644   .0244378   -19.63   0.000    -.5275412   -.4317468
               Ticker_num |  -.0003272   .0002069    -1.58   0.114    -.0007326    .0000783
             Currency_num |   .0267972   .0033955     7.89   0.000     .0201422    .0334523
          PaymentRank_num |   .1183414   .0266971     4.43   0.000     .0660161    .1706666
           MaturityYears2 |   .0067621   .0062303     1.09   0.278    -.0054491    .0189732
             AmountIssued |   1.24e-13   5.09e-14     2.44   0.015     2.44e-14    2.24e-13
                    _cons |   .4719899   .2108222     2.24   0.025      .058786    .8851937
--------------------------+----------------------------------------------------------------
 var(e.YieldtoMaturityMid)|   13.38312   .1950987                      13.00614    13.77102
-------------------------------------------------------------------------------------------

[Joseph Coveney]

I don't know what a Mundlak model is, but you didn't specify any random effects (clustering variable), and so you just fitted a conventional linear regression model.

As far as goodness of fit, you could try examining predictions graphically.