Statalist


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

st: Xtpoisson, fe bug?


From   "Mark Bryan" <[email protected]>
To   <[email protected]>
Subject   st: Xtpoisson, fe bug?
Date   Tue, 22 Apr 2008 11:50:53 +0100

Dear Statalist,

I run the following fixed effects Poisson model:

. version 10.0
. query born
03 Mar 2008

. xtpoisson norgm age agesq cohort female married ///
> further degree emp pt preschkid schkid, fe
note: 3326 groups (3326 obs) dropped because of only one obs per group
note: 2473 groups (12215 obs) dropped because of all zero outcomes
note: cohort omitted because it is constant within group
note: female omitted because it is constant within group

<output omitted>

Conditional fixed-effects Poisson regression    Number of obs      =
64987
Group variable: pid                             Number of groups   =
9681

                                                Obs per group: min =
2
                                                               avg =
6.7
                                                               max =
9

                                                Wald chi2(9)       =
179.56
Log likelihood  = -53509.648                    Prob > chi2        =
0.0000

----------------------------------------------------------------------------
--
       norgm |      Coef.   Std. Err.      z    P>|z|     [95% Conf.
Interval]
-------------+--------------------------------------------------------------
--
         age |   .0281423   .0036906     7.63   0.000     .0209089
.0353758
       agesq |  -.0003175   .0000361    -8.78   0.000    -.0003883
-.0002467
      cohort |   .0257127   .0185708     1.38   0.166    -.0106854
.0621108
      female |  -.0117726   .0237682    -0.50   0.620    -.0583574
.0348123
     married |  -.0664054    .039699    -1.67   0.094     -.144214
.0114032
     further |   .0539167   .0150458     3.58   0.000     .0244275
.0834059
      degree |   -.002381   .0171409    -0.14   0.890    -.0359765
.0312145
         emp |  -.0557617   .0167727    -3.32   0.001    -.0886355
-.0228878
          pt |   .0634887   .0141329     4.49   0.000     .0357888
.0911886
----------------------------------------------------------------------------
--

Stata correctly identifies that cohort and female are time-invariant and
drops them. 
However, they appear in the output, while preschkid and schkid are left out.

I rerun the model omitting cohort and female from the variable list:

. xtpoisson norgm age agesq /*cohort female*/ married ///
> further degree emp pt preschkid schkid, fe
note: 3326 groups (3326 obs) dropped because of only one obs per group
note: 2473 groups (12215 obs) dropped because of all zero outcomes

< output omitted >

Conditional fixed-effects Poisson regression    Number of obs      =
64987
Group variable: pid                             Number of groups   =
9681

                                                Obs per group: min =
2
                                                               avg =
6.7
                                                               max =
9

                                                Wald chi2(9)       =
179.56
Log likelihood  = -53509.648                    Prob > chi2        =
0.0000

----------------------------------------------------------------------------
--
       norgm |      Coef.   Std. Err.      z    P>|z|     [95% Conf.
Interval]
-------------+--------------------------------------------------------------
--
         age |   .0281423   .0036906     7.63   0.000     .0209089
.0353758
       agesq |  -.0003175   .0000361    -8.78   0.000    -.0003883
-.0002467
     married |   .0257127   .0185708     1.38   0.166    -.0106854
.0621108
     further |  -.0117726   .0237682    -0.50   0.620    -.0583574
.0348123
      degree |  -.0664054    .039699    -1.67   0.094     -.144214
.0114032
         emp |   .0539167   .0150458     3.58   0.000     .0244275
.0834059
          pt |   -.002381   .0171409    -0.14   0.890    -.0359765
.0312145
   preschkid |  -.0557617   .0167727    -3.32   0.001    -.0886355
-.0228878
      schkid |   .0634887   .0141329     4.49   0.000     .0357888
.0911886
----------------------------------------------------------------------------
-- 

Results are identical except the variable labelling now seems correct
(preschkid and schkid now included).
Looks like a bug?

Mark


----------------------
Mark Bryan ([email protected])
Institute for Social and Economic Research (ISER)
University of Essex, Colchester, UK CO4 3SQ
tel: +44 (0)1206 874683; fax: +44 (0)1206 873151


 


*
*   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