logistic regression

[Nermine Medhioub]

Hello everyone, I have a problem with the panel logistic regression method. When I ran xtlogit with both random and fixed effects, I got this result and it never ends
Iteration 0: log likelihood = -4.459e-13 (not concave)
Iteration 1: log likelihood = -4.285e-13 (not concave)
Iteration 2: log likelihood = -1.990e-13 (not concave)
Iteration 3: log likelihood = -1.990e-13 (not concave)
Iteration 4: log likelihood = -1.990e-13 (not concave)
Iteration 5: log likelihood = -1.990e-13 (not concave)
Iteration 6: log likelihood = -1.990e-13 (not concave)
Iteration 7: log likelihood = -1.990e-13 (not concave)
Iteration 8: log likelihood = -1.990e-13 (not concave)
Iteration 9: log likelihood = -1.990e-13 (not concave)
Iteration 10: log likelihood = -1.990e-13 (not concave)
Iteration 11: log likelihood = -1.990e-13 (not concave)
Iteration 12: log likelihood = -1.990e-13 (not concave)
Iteration 13: log likelihood = -1.990e-13 (not concave)
Iteration 14: log likelihood = -1.990e-13 (not concave)
Iteration 15: log likelihood = -1.990e-13 (not concave)
Iteration 16: log likelihood = -1.990e-13 (not concave)
Iteration 17: log likelihood = -1.990e-13 (not concave)
Iteration 18: log likelihood = -1.990e-13 (not concave)
Iteration 19: log likelihood = -1.990e-13 (not concave)
Iteration 20: log likelihood = -1.990e-13 (not concave)
Iteration 21: log likelihood = -1.990e-13 (not concave)
Iteration 22: log likelihood = -1.990e-13 (not concave)
Iteration 23: log likelihood = -1.990e-13 (not concave)
Iteration 24: log likelihood = -1.990e-13 (not concave)
Iteration 25: log likelihood = -1.990e-13 (not concave)
Iteration 26: log likelihood = -1.990e-13 (not concave)
Iteration 27: log likelihood = -1.990e-13 (not concave)
Iteration 28: log likelihood = -1.990e-13 (not concave)
Iteration 29: log likelihood = -1.990e-13 (not concave)
Iteration 30: log likelihood = -1.990e-13 (not concave)
Iteration 31: log likelihood = -1.990e-13 (not concave)
Iteration 32: log likelihood = -1.990e-13 (not concave)
Iteration 33: log likelihood = -1.990e-13 (not concave)
Iteration 34: log likelihood = -1.990e-13 (not concave)
Iteration 35: log likelihood = -1.990e-13 (not concave)

[Carlo Lazzaro]

Nermine:
try with a more parsimonius model (that is, less predictors).
Then add one predictor at a time and see when Stata starts to complain.

[Nermine Medhioub]

sorry @Carlo Lazzaro I didn't understand what you mean

[Carlo Lazzaro]

Nermine:
the first code of the following toy-example makes the -logit- machinery gasping (##lenght is the culprit).
Therefore, I started it all over again with a simple -logit- and then investigated possible convergence issues adding -weight- until its 4 power:

. logit foreign c.weight##c.weight##c.weight## length

<snip>
Iteration 0:  Log likelihood = -6.1826542  
Iteration 1:  Log likelihood = -1.4547199  (not concave)
Iteration 2:  Log likelihood = -1.4106456  (not concave)
Iteration 3:  Log likelihood = -1.4078625  (not concave)
Iteration 4:  Log likelihood = -1.4075116  (not concave)
Iteration 5:  Log likelihood = -1.4068567  (not concave)
Iteration 6:  Log likelihood = -1.4066832  (not concave)
Iteration 7:  Log likelihood = -1.4065951  (not concave)
Iteration 8:  Log likelihood = -1.4065216  (not concave)
Iteration 9:  Log likelihood = -1.4064535  (not concave)
Iteration 10: Log likelihood = -1.4063855  (not concave)
Iteration 11: Log likelihood = -1.4063226  (not concave)
Iteration 12: Log likelihood =  -1.406264  (not concave)
Iteration 13: Log likelihood = -1.4062073  (not concave)
Iteration 14: Log likelihood = -1.4061534  (not concave)
Iteration 15: Log likelihood = -1.4061022  (not concave)
Iteration 16: Log likelihood = -1.4060534  (not concave)
Iteration 17: Log likelihood = -1.4060067  (not concave)
Iteration 18: Log likelihood = -1.4059622  (not concave)
Iteration 19: Log likelihood = -1.4059197  (not concave)
Iteration 20: Log likelihood = -1.4058791  (not concave)
Iteration 21: Log likelihood = -1.4058403  (not concave)
Iteration 22: Log likelihood = -1.4058032  (not concave)
Iteration 23: Log likelihood = -1.4057677  (not concave)
Iteration 24: Log likelihood = -1.4057337  (not concave)
Iteration 25: Log likelihood = -1.4057012  (not concave)
Iteration 26: Log likelihood = -1.4056701  (not concave)
Iteration 27: Log likelihood = -1.4056403  (not concave)
Iteration 28: Log likelihood = -1.4056117  (not concave)
Iteration 29: Log likelihood = -1.4055843  (not concave)
Iteration 30: Log likelihood = -1.4055581  (not concave)
Iteration 31: Log likelihood = -1.4055329  (not concave)
Iteration 32: Log likelihood = -1.4055087  (not concave)
Iteration 33: Log likelihood = -1.4054855  (not concave)
Iteration 34: Log likelihood = -1.4054632  (not concave)
Iteration 35: Log likelihood = -1.4054418  (not concave)
Iteration 36: Log likelihood = -1.4054213  (not concave)
Iteration 37: Log likelihood = -1.4054015  (not concave)
Iteration 38: Log likelihood = -1.4053825  (not concave)
Iteration 39: Log likelihood = -1.4053642  (not concave)

. logit foreign c.weight

Iteration 0:  Log likelihood =  -45.03321  
Iteration 1:  Log likelihood = -30.669507  
Iteration 2:  Log likelihood = -29.068209  
Iteration 3:  Log likelihood = -29.054005  
Iteration 4:  Log likelihood = -29.054002  
Iteration 5:  Log likelihood = -29.054002  

Logistic regression                                     Number of obs =     74
                                                        LR chi2(1)    =  31.96
                                                        Prob > chi2   = 0.0000
Log likelihood = -29.054002                             Pseudo R2     = 0.3548

------------------------------------------------------------------------------
     foreign | Coefficient  Std. err.      z    P>|z|     [95% conf. interval]
-------------+----------------------------------------------------------------
      weight |  -.0025874   .0006094    -4.25   0.000    -.0037817    -.001393
       _cons |   6.282599   1.603967     3.92   0.000     3.138882    9.426316
------------------------------------------------------------------------------

. logit foreign c.weight##c.weight

Iteration 0:  Log likelihood =  -45.03321  
Iteration 1:  Log likelihood = -30.804791  
Iteration 2:  Log likelihood = -29.098259  
Iteration 3:  Log likelihood = -28.482097  
Iteration 4:  Log likelihood = -28.449536  
Iteration 5:  Log likelihood = -28.449396  
Iteration 6:  Log likelihood = -28.449396  

Logistic regression                                     Number of obs =     74
                                                        LR chi2(2)    =  33.17
                                                        Prob > chi2   = 0.0000
Log likelihood = -28.449396                             Pseudo R2     = 0.3683

-----------------------------------------------------------------------------------
          foreign | Coefficient  Std. err.      z    P>|z|     [95% conf. interval]
------------------+----------------------------------------------------------------
           weight |   .0045252   .0068448     0.66   0.509    -.0088905    .0179408
                  |
c.weight#c.weight |  -1.33e-06   1.30e-06    -1.02   0.307    -3.88e-06    1.22e-06
                  |
            _cons |  -2.853664   8.752615    -0.33   0.744    -20.00847    14.30115
-----------------------------------------------------------------------------------

. logit foreign c.weight##c.weight##c.weight

Iteration 0:  Log likelihood =  -45.03321  
Iteration 1:  Log likelihood = -30.300263  
Iteration 2:  Log likelihood = -28.653719  
Iteration 3:  Log likelihood = -28.438623  
Iteration 4:  Log likelihood = -28.405448  
Iteration 5:  Log likelihood = -28.401876  
Iteration 6:  Log likelihood = -28.401847  
Iteration 7:  Log likelihood = -28.401847  

Logistic regression                                     Number of obs =     74
                                                        LR chi2(2)    =  33.26
                                                        Prob > chi2   = 0.0000
Log likelihood = -28.401847                             Pseudo R2     = 0.3693

--------------------------------------------------------------------------------------------
                   foreign | Coefficient  Std. err.      z    P>|z|     [95% conf. interval]
---------------------------+----------------------------------------------------------------
                    weight |   .0169144   .0365587     0.46   0.644    -.0547394    .0885682
                           |
         c.weight#c.weight |  -6.04e-06   .0000135    -0.45   0.656    -.0000326    .0000205
                           |
c.weight#c.weight#c.weight |   5.80e-10   1.64e-09     0.35   0.723    -2.63e-09    3.79e-09
                           |
                     _cons |  -13.43712   32.16286    -0.42   0.676    -76.47517    49.60093
--------------------------------------------------------------------------------------------

. logit foreign c.weight##c.weight##c.weight##c.weight

Iteration 0:  Log likelihood =  -45.03321  
Iteration 1:  Log likelihood = -30.015321  
Iteration 2:  Log likelihood = -28.589222  
Iteration 3:  Log likelihood = -28.354432  
Iteration 4:  Log likelihood = -28.101003  
Iteration 5:  Log likelihood = -27.871035  
Iteration 6:  Log likelihood = -27.814182  
Iteration 7:  Log likelihood = -27.812103  
Iteration 8:  Log likelihood = -27.812099  
Iteration 9:  Log likelihood = -27.812099  

Logistic regression                                     Number of obs =     74
                                                        LR chi2(2)    =  34.44
                                                        Prob > chi2   = 0.0000
Log likelihood = -27.812099                             Pseudo R2     = 0.3824

-----------------------------------------------------------------------------------------------------
                            foreign | Coefficient  Std. err.      z    P>|z|     [95% conf. interval]
------------------------------------+----------------------------------------------------------------
                             weight |   .3710759   .3788477     0.98   0.327     -.371452    1.113604
                                    |
                  c.weight#c.weight |  -.0002112   .0002234    -0.95   0.345    -.0006491    .0002267
                                    |
         c.weight#c.weight#c.weight |   5.24e-08   5.77e-08     0.91   0.364    -6.08e-08    1.66e-07
                                    |
c.weight#c.weight#c.weight#c.weight |  -4.82e-12   5.52e-12    -0.87   0.382    -1.56e-11    6.00e-12
                                    |
                              _cons |  -238.2601   237.3854    -1.00   0.316    -703.5269    227.0066
-----------------------------------------------------------------------------------------------------
Note: 5 failures and 0 successes completely determined.

.