Re: st: c_logit and dropped observations

["Arne Risa Hole" <arnehole@gmail.com>]

Hi Rodrigo

This problem is probably caused by one or more of your independent
variables having missing values. If one of the independent variables
is missing this observation will be dropped; if the dropped
observation is the one containing the chosen alternative (i.e. the one
in which the dependent variable is one) the dependent variable is zero
for the remaining (included) observations within the group.

This also explains why Stata reports that 1,446 observations are
dropped due to all positive or all negative outcomes since it doesn't
count the 6 observations that are dropped due to the missing
independent variable(s) (1,452-6=1446).

Hope this helps.

Arne


On 01/10/06, R.Alegria@lse.ac.uk <R.Alegria@lse.ac.uk> wrote:
> Dear Satalist,
>
> I am estimating a conditional logit with the following characteristics:
>
> -Number of decision makers=6,329
> -Number of choices=242
> -Number of observations=1,531,618 (=6,329*242)
> -Decision makers are firms which decide where to locate over a set of 242 location alternatives and independent variables are economic characteristics of the location alternatives.
>
> When estimating the conditional logit I obtain the following message:
>
> "6 groups (1446 obs) dropped due to all positive or all negative outcomes"
>
> clogit choice  n2_gdp n2_gdppc n2_mp n2_w2_2 n2_unemp_r l_n2_managg_s n2_fdi n2_popden, group(id)
> note: 6 groups (1446 obs) dropped due to all positive or
>       all negative outcomes.
> Iteration 0:   log likelihood = -31564.979
> Iteration 1:   log likelihood = -31562.865
> Iteration 2:   log likelihood = -31562.865
> Conditional (fixed-effects) logistic regression   Number of obs   =    1529212
>                                                   LR chi2(8)      =    6279.48
>                                                   Prob > chi2     =     0.0000
> Log likelihood = -31562.865                       Pseudo R2       =     0.0905
> ------------------------------------------------------------------------------
>       choice |      Coef.   Std. Err.      z    P>|z|     [95% Conf. Interval]
> -------------+----------------------------------------------------------------
>       n2_gdp |   .2244931   .0260914     8.60   0.000     .1733548    .2756314
>     n2_gdppc |  -.3560138    .033739   -10.55   0.000    -.4221411   -.2898865
>        n2_mp |   .0513986   .0350509     1.47   0.143    -.0172999    .1200972
>      n2_w2_2 |  -.1733384    .047729    -3.63   0.000    -.2668855   -.0797913
>   n2_unemp_r |  -.0770995   .0209851    -3.67   0.000    -.1182295   -.0359696
> l_n2_manag~s |   .0490496   .0404983     1.21   0.226    -.0303257    .1284248
>       n2_fdi |   .9064167   .0160668    56.42   0.000     .8749263     .937907
>    n2_popden |   .4422606   .0792087     5.58   0.000     .2870145    .5975067
> ------------------------------------------------------------------------------
>
>
> As far as I know this means that within those 6 groups (in my case firms) all observations have the same value for the dependent variable.
>
> However I have made several checks and I have one 1 for each group and the rest zeros and also have checked that there are 242 choices for each group.
>
> 1/ Can anyone tell me why I am still having this problem?
>
> 2/ Is there any way to say stata to tell me which are the 6 groups which are being dropped?
>
> 3/ If 6 groups are being dropped, then 6*242=1,452 observations should be dropped but in the message above the observations dropped are 1,446. Why?
>
> I appreciate any comment or help,
>
> Thanks!,
>
> ---Rodrigo
>
>
>
>
> *
> *   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/