Statalist


[Date Prev][Date Next][Thread Prev][Thread Next][Date index][Thread index]

st: xttobit: Different versions, different results?


From   Nikos Nikiforakis <n.nikiforakis@unimelb.edu.au>
To   statalist@hsphsun2.harvard.edu
Subject   st: xttobit: Different versions, different results?
Date   Fri, 05 Sep 2008 13:41:28 +1000

Dear Statalist users,
 
I have trouble replicating results I obtained with xttobit in previous versions of Stata. In particular, xttobit gives substantially different results in versions 8, 9, and 10 despite using the same set of variables and the same dataset.
 
Below you can see the output from each version. Does anyone know why this is happening or encountered a similar problem?
 
VERSION 8:
(unfortunately, I no longer have the information about the iteration for this version)
 
. xttobit mycont three asym_level strong, i(group) ul(20) ll(0)
Random-effects tobit regression                 Number of obs      =      1320
Group variable (i): group                       Number of groups   =        33
Random effects u_i ~ Gaussian                   Obs per group: min =        40
                                                               avg =      40.0
                                                               max =        40
                                                Wald chi2(3)       =     38.02
Log likelihood  = -3000.3546                    Prob > chi2        =    0.0000
------------------------------------------------------------------------------
      mycont |      Coef.   Std. Err.      z    P>|z|     [95% Conf. Interval]
-------------+----------------------------------------------------------------
       three |   3.144437   .5602632     5.61   0.000     2.046342    4.242533
  asym_level |   .1939476   .2007664     0.97   0.334    -.1995474    .5874425
      strong |  -1.369307   .5868873    -2.33   0.020    -2.519585   -.2190286
       _cons |   11.52007   .6803106    16.93   0.000     10.18668    12.85345
-------------+----------------------------------------------------------------
    /sigma_u |   7.943933   1.000435     7.94   0.000     5.983116    9.904751
    /sigma_e |    6.75084   .1805469    37.39   0.000     6.396975    7.104705
-------------+----------------------------------------------------------------
         rho |     .58066   .0621167                      .4570053    .6967743
------------------------------------------------------------------------------
  Observation summary:        58  left-censored observations
                             774     uncensored observations
                             488 right-censored observations
VERSION 9.2:
. xttobit mycont asym_level strong three, ll(0) ul(20)  i(group)
Obtaining starting values for full model:
Iteration 0:   log likelihood = -3918.8279
Iteration 1:   log likelihood = -3918.8114
Iteration 2:   log likelihood = -3918.8114
Fitting full model:
Iteration 0:   log likelihood = -3113.7583 
Iteration 1:   log likelihood =  -2977.791 
Iteration 2:   log likelihood =  -2974.585 
Iteration 3:   log likelihood = -2974.5706 
Iteration 4:   log likelihood = -2974.5706 
Random-effects tobit regression                 Number of obs      =      1320
Group variable: group                           Number of groups   =        33
Random effects u_i ~ Gaussian                   Obs per group: min =        40
                                                               avg =      40.0
                                                               max =        40
                                                Wald chi2(3)       =      7.82
Log likelihood  = -2974.5706                    Prob > chi2        =    0.0498
------------------------------------------------------------------------------
      mycont |      Coef.   Std. Err.      z    P>|z|     [95% Conf. Interval]
-------------+----------------------------------------------------------------
  asym_level |    .082819   .7468687     0.11   0.912    -1.381017    1.546655
      strong |  -1.312225   .5864039    -2.24   0.025    -2.461555   -.1628941
       three |   3.021097   1.854002     1.63   0.103    -.6126807    6.654875
       _cons |   13.92143   2.425767     5.74   0.000     9.167013    18.67585
-------------+----------------------------------------------------------------
    /sigma_u |   7.531323   .9907186     7.60   0.000      5.58955    9.473096
    /sigma_e |   6.668207   .1836982    36.30   0.000     6.308165    7.028249
-------------+----------------------------------------------------------------
         rho |   .5605611   .0652327                      .4317845    .6831824
------------------------------------------------------------------------------
  Observation summary:        58  left-censored observations
                             774     uncensored observations
                             488 right-censored observations
 
VERSION 10:
. xttobit mycont asym_level strong three, ll(0) ul(20)  i(group)
Obtaining starting values for full model:
Iteration 0:   log likelihood = -3918.8279
Iteration 1:   log likelihood = -3918.8114
Iteration 2:   log likelihood = -3918.8114
Fitting full model:
Iteration 0:   log likelihood = -3112.6087 
Iteration 1:   log likelihood = -2976.4144 
Iteration 2:   log likelihood = -2974.0119 
Iteration 3:   log likelihood = -2973.9993 
Iteration 4:   log likelihood = -2973.9993 
Random-effects tobit regression                 Number of obs      =      1320
Group variable: group                           Number of groups   =        33
Random effects u_i ~ Gaussian                   Obs per group: min =        40
                                                               avg =      40.0
                                                               max =        40
                                                Wald chi2(3)       =      8.47
Log likelihood  = -2973.9993                    Prob > chi2        =    0.0372
------------------------------------------------------------------------------
      mycont |      Coef.   Std. Err.      z    P>|z|     [95% Conf. Interval]
-------------+----------------------------------------------------------------
  asym_level |   -.080391   .9446354    -0.09   0.932    -1.931842     1.77106
      strong |  -1.291137   .5891479    -2.19   0.028    -2.445846   -.1364281
       three |   4.746394   2.791425     1.70   0.089    -.7246978    10.21749
       _cons |   14.47964   3.339526     4.34   0.000     7.934287    21.02499
-------------+----------------------------------------------------------------
    /sigma_u |   7.178246   .9502204     7.55   0.000     5.315848    9.040644
    /sigma_e |   6.683428   .1857913    35.97   0.000     6.319284    7.047573
-------------+----------------------------------------------------------------
         rho |   .5356514    .066032                      .4066286    .6610002
------------------------------------------------------------------------------
  Observation summary:        58  left-censored observations
                             774     uncensored observations
                             488 right-censored observations
 
I have seen one previous post on the topic (http://www.stata.com/support/faqs/stat/xttobit.html). The explanation however does not seem to apply to my problem. (Note that I did not have the same problem using xtreg or xtprobit).
 
Thank you in advance for your responses.
 
Nikos Nikiforakis
 
Nikos Nikiforakis
Dept. Economics
The University of Melbourne
http://www.economics.unimelb.edu.au/nnikiforakis/



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