Store significant variables in Logistic Regression

[Diego Saverys]

Hi,

I am running a logistic regression with about 30 variables, most of which are categorical variables, and I would like to run a command afterwards that creates a list containing all the significant variables (say at the 5% level) to run some further analysis on those specific variables.

Do you have any idea on how to do this ?

Thanks,
Diego

[Maarten Buis]

. sysuse auto, clear
(1978 automobile data)

. // I am deliberately using a model with ommited
. // variables, empty cells, adding the same variable twice,
. // to see if my solution works under difficult circumstances
. logit foreign i.rep78 mpg mpg weight length, or

note: 1.rep78 != 0 predicts failure perfectly;
      1.rep78 omitted and 2 obs not used.

note: 2.rep78 != 0 predicts failure perfectly;
      2.rep78 omitted and 8 obs not used.

note: 5.rep78 omitted because of collinearity.
note: mpg omitted because of collinearity.
Iteration 0:   log likelihood = -38.411464  
Iteration 1:   log likelihood = -18.363164  
Iteration 2:   log likelihood = -16.168782  
Iteration 3:   log likelihood = -15.891852  
Iteration 4:   log likelihood = -15.889134  
Iteration 5:   log likelihood = -15.889132  
Iteration 6:   log likelihood = -15.889132  

Logistic regression                                     Number of obs =     59
                                                        LR chi2(5)    =  45.04
                                                        Prob > chi2   = 0.0000
Log likelihood = -15.889132                             Pseudo R2     = 0.5863

------------------------------------------------------------------------------
     foreign | Odds ratio   Std. err.      z    P>|z|     [95% conf. interval]
-------------+----------------------------------------------------------------
       rep78 |
          1  |          1  (empty)
          2  |          1  (empty)
          3  |   .0275865    .038135    -2.60   0.009     .0018366    .4143617
          4  |   .2909225   .3771016    -0.95   0.341     .0229312    3.690861
          5  |          1  (omitted)
             |
         mpg |   .7868654   .0960931    -1.96   0.050     .6193704    .9996556
         mpg |          1  (omitted)
      weight |    .994905   .0028337    -1.79   0.073     .9893666    1.000474
      length |   1.033117   .0850567     0.40   0.692     .8791641    1.214029
       _cons |    2469905   2.56e+07     1.42   0.155     .0037979    1.61e+15
------------------------------------------------------------------------------
Note: _cons estimates baseline odds.

.
.
. local k = colsof(e(b))

. local x : colnames(e(b))

.
. tempname table

. matrix `table' = r(table)

.
. forvalues i = 1/`k' {
  2.    if `table'[4,`i'] < 0.05 {        
  3.        local res = `"`res' `:word `i' of `x''"'
  4.     }
  5. }

. di "`res'"
 3.rep78 mpg