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

Re: st: xtprobit vs. probit


From   Christopher Auld <auld@ucalgary.ca>
To   statalist@hsphsun2.harvard.edu
Subject   Re: st: xtprobit vs. probit
Date   Thu, 25 Jul 2002 09:20:11 -0600 (MDT)

It isn't that the estimated model's rho=0, rather, the panel model failed
to converge.  As a rule, whenever a standard error is missing, something
is seriously wrong (in my humble opinion, Stata should flag the problem
more prominently than that implied by failing to report standard errors,
like "WARNING: Hessian is singular.  Results are meaningless.").

Why the model didn't converge is another question. I'd hazard a guess that
if you have too many units which are all 0 or all 1 the panel probit model
will have difficulties.


Cheers,

Chris Auld
Department of Economics
University of Calgary
auld@ucalgary.ca

On Thu, 25 Jul 2002, Davide Castellani wrote:

>
> Dear statalisters,
> I am trying to run a random effect probit using -xtprobit, re-, but it
> yields the same results as the standard -probit- command. Furthermore, the
> xtprobit yields strange
> output for lnsig2u, sigma_u and rho.  The same does not happen if I use
> xtlogit. Below i report output for the xtprobit, probit, logit and xtlogit.
> Do you have any idea of what could be going on?
>
> Davide
>
>
> . xtprobit dep x1 x2 if time<1998, re i(id)
>
> Fitting comparison model:
>
> Iteration 0:   log likelihood = -1611.7984
> Iteration 1:   log likelihood = -1371.4379
> Iteration 2:   log likelihood = -1361.8071
> Iteration 3:   log likelihood = -1361.7691
>
> Fitting full model:
>
> rho =  0.0     log likelihood = -1361.7691
> rho =  0.1     log likelihood = -1362.7407
> Iteration 0:   log likelihood = -1361.7691
>
> Random-effects probit                           Number of obs      =      9001
> Group variable (i) : id                         Number of groups   =      2602
>
> Random effects u_i ~ Gaussian                   Obs per group: min =         1
>                                                                 avg =       3.5
>                                                                 max =         6
>
>                                                  Wald chi2(2)       =    467.84
> Log likelihood  = -1361.7691                    Prob > chi2        =    0.0000
>
> ------------------------------------------------------------------------------
>           dep |      Coef.   Std. Err.      z    P>|z|     [95% Conf. Interval]
> -------------+----------------------------------------------------------------
>            x1 |   .5067719   .0560958     9.03   0.000     .3968261    .6167178
>            x2 |    .327031   .0186308    17.55   0.000     .2905153    .3635466
>         _cons |   -3.31608   .0953378   -34.78   0.000    -3.502938   -3.129221
> -------------+----------------------------------------------------------------
>      /lnsig2u |        -14          .                             .           .
> -------------+----------------------------------------------------------------
>       sigma_u |   .0009119          .                             .           .
>           rho |   8.32e-07          .                             .           .
> ------------------------------------------------------------------------------
> Likelihood ratio test of rho=0: chibar2(01) =     0.00 Prob >= chibar2 = 1.000
>
> . probit dep x1 x2 if time<1998
>
> Iteration 0:   log likelihood = -1611.7984
> Iteration 1:   log likelihood = -1371.4379
> Iteration 2:   log likelihood = -1361.8071
> Iteration 3:   log likelihood = -1361.7691
>
> Probit estimates                                  Number of obs   =       9001
>                                                    LR chi2(2)      =     500.06
>                                                    Prob > chi2     =     0.0000
> Log likelihood = -1361.7691                       Pseudo R2       =     0.1551
>
> ------------------------------------------------------------------------------
>           dep |      Coef.   Std. Err.      z    P>|z|     [95% Conf. Interval]
> -------------+----------------------------------------------------------------
>            x1 |   .5067717   .0560951     9.03   0.000     .3968274     .616716
>            x2 |   .3270309   .0186305    17.55   0.000     .2905158    .3635459
>         _cons |  -3.316078   .0953354   -34.78   0.000    -3.502932   -3.129224
> ------------------------------------------------------------------------------
>
>
>
> . logit dep x1 x2 if time<1998
>
> Iteration 0:   log likelihood = -1611.7984
> Iteration 1:   log likelihood = -1587.3111
> Iteration 2:   log likelihood = -1373.1369
> Iteration 3:   log likelihood = -1369.2397
> Iteration 4:   log likelihood = -1369.2142
>
> Logit estimates                                   Number of obs   =       9001
>                                                    LR chi2(2)      =     485.17
>                                                    Prob > chi2     =     0.0000
> Log likelihood = -1369.2142                       Pseudo R2       =     0.1505
>
> ------------------------------------------------------------------------------
>           dep |      Coef.   Std. Err.      z    P>|z|     [95% Conf. Interval]
> -------------+----------------------------------------------------------------
>            x1 |   .9555873   .1115672     8.57   0.000     .7369195    1.174255
>            x2 |   .6487152   .0359413    18.05   0.000     .5782716    .7191588
>         _cons |  -6.320212   .1958625   -32.27   0.000    -6.704095   -5.936329
> ------------------------------------------------------------------------------
>
> . xtlogit dep x1 x2 if time<1998, re i(id)
>
> Fitting comparison model:
>
> Iteration 0:   log likelihood = -1611.7984
> Iteration 1:   log likelihood = -1587.3111
> Iteration 2:   log likelihood = -1373.1369
> Iteration 3:   log likelihood = -1369.2397
> Iteration 4:   log likelihood = -1369.2142
>
> Fitting full model:
>
> rho =  0.0     log likelihood = -1369.2142
> rho =  0.1     log likelihood = -1367.2316
> rho =  0.2     log likelihood = -1366.5762
> rho =  0.3     log likelihood = -1368.2915
> Iteration 0:   log likelihood = -1366.5762
> Iteration 1:   log likelihood = -1363.6172
> Iteration 2:   log likelihood = -1362.5641
> Iteration 3:   log likelihood = -1362.5623
> Iteration 4:   log likelihood = -1362.5623
>
> Random-effects logit                            Number of obs      =      9001
> Group variable (i) : id                         Number of groups   =      2602
>
> Random effects u_i ~ Gaussian                   Obs per group: min =         1
>                                                                 avg =       3.5
>                                                                 max =         6
>
>                                                  Wald chi2(2)       =    341.10
> Log likelihood  = -1362.5623                    Prob > chi2        =    0.0000
>
> ------------------------------------------------------------------------------
>           dep |      Coef.   Std. Err.      z    P>|z|     [95% Conf. Interval]
> -------------+----------------------------------------------------------------
>            x1 |   1.040364   .1279023     8.13   0.000     .7896805    1.291048
>            x2 |   .6961738   .0447027    15.57   0.000     .6085581    .7837896
>         _cons |  -6.843286   .2759503   -24.80   0.000    -7.384139   -6.302434
> -------------+----------------------------------------------------------------
>      /lnsig2u |  -.4303919   .3243294                     -1.066066    .2052821
> -------------+----------------------------------------------------------------
>       sigma_u |   .8063834   .1307669                      .5868225    1.108094
>           rho |    .165034   .0135847                      .0947548    .2717889
> ------------------------------------------------------------------------------
> Likelihood ratio test of rho=0: chibar2(01) =    13.30 Prob >= chibar2 = 0.000
>
> .
> end of do-file
>
> *
> *   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/
>

*
*   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 1996–2014 StataCorp LP   |   Terms of use   |   Privacy   |   Contact us   |   What's new   |   Site index