Stata The Stata listserver
[Date Prev][Date Next][Thread Prev][Thread Next][Date index][Thread index]

st: RE: xtlogit (re) and logit: same results


From   "Andreas Kamp" <[email protected]>
To   <[email protected]>
Subject   st: RE: xtlogit (re) and logit: same results
Date   Tue, 19 Jul 2005 10:37:12 +0100

Dear Stata-Users,

as I still haven't found an answer to my question, let me try to precise
my question. 

As I mentioned before, I am trying to estimate default probabilities
using a pooled logit model (logit) and a random effects panel logit
model (xtlogit, re). Results for both models are perfectley the same.

Here are the results for xtlogit, re:

Fitting comparison model:

Iteration 0:   log likelihood =  -1377.398
Iteration 1:   log likelihood = -1342.3038
Iteration 2:   log likelihood = -1329.8824
Iteration 3:   log likelihood = -1328.7468
Iteration 4:   log likelihood = -1328.7417
Iteration 5:   log likelihood = -1328.7417

Fitting full model:

tau =  0.0     log likelihood = -1328.7417
tau =  0.1     log likelihood = -1330.1846

Iteration 0:   log likelihood = -1328.7417  
Iteration 1:   log likelihood = -1328.7417  

Random-effects logistic regression              Number of obs      =
19895
Group variable (i): sysnr                       Number of groups   =
3164

Random effects u_i ~ Gaussian                   Obs per group: min =
1
                                                               avg =
6.3
                                                               max =
8

                                                Wald chi2(7)       =
101.89
Log likelihood  = -1328.7417                    Prob > chi2        =
0.0000

------------------------------------------------------------------------
------
     ausfall |      Coef.   Std. Err.      z    P>|z|     [95% Conf.
Interval]
-------------+----------------------------------------------------------
------
      mas1_2 |  -1.363757   .6244236    -2.18   0.029    -2.587605
-.1399094
    ejb128_2 |  -.2314217   .0711395    -3.25   0.001    -.3708526
-.0919908
       ekq_2 |  -18.38955    6.88451    -2.67   0.008    -31.88294
-4.896155
       gkr_2 |  -129.3035   26.23809    -4.93   0.000    -180.7292
-77.87775
        pi_2 |  -1.447272   .6023384    -2.40   0.016    -2.627833
-.2667099
      notl_2 |   2.162819   .3786747     5.71   0.000      1.42063
2.905008
        geno |   .5634791   .2215662     2.54   0.011     .1292174
.9977408
       _cons |   .9511411   1.598035     0.60   0.552    -2.180949
4.083232
-------------+----------------------------------------------------------
------
    /lnsig2u |  -14.99999   585.2423                     -1162.054
1132.054
-------------+----------------------------------------------------------
------
     sigma_u |   .0005531   .1618446                      4.6e-253
6.6e+245
         rho |   9.30e-08   .0000544                             0
.
------------------------------------------------------------------------
------
Likelihood-ratio test of rho=0: chibar2(01) =     0.00 Prob >= chibar2 =
1.000

I read somewhere that a model might not converge because it has too few
defaults (dependend variable = 1). However, I don't understand this
argument. Why does the model not converge if it has to few observations
equalt to 1? 

Thank you for your help,

Andreas 


*
*   For searches and help try:
*   http://www.stata.com/support/faqs/res/findit.html
*   http://www.stata.com/support/statalist/faq
*   http://www.ats.ucla.edu/stat/stata/



© Copyright 1996–2024 StataCorp LLC   |   Terms of use   |   Privacy   |   Contact us   |   What's new   |   Site index