Multivariable logistic regression missing out categories

[Josie Thompson]

Hi all,
I'm trying to a multivariable logistic regression analysis of some data, with one of the variables being 'period'

* Example generated by -dataex-. For more info, type help dataex
clear
input float period
1
1
1
1
1
1
1
1
6
2
6
3
6
6
3
1
1
1
1
1
1
2
6
1
2
3
6
4
1
5
1
1
1
1
1
1
1
1
3
7
6
7
7
6
5
1
1
6
6
1
1
2
6
1
6
1
4
3
6
3
3
7
1
6
6
4
2
1
3
6
3
1
1
1
6
3
2
7
1
7
6
3
7
6
2
6
6
1
6
6
6
3
7
6
1
6
3
6
6
6
end
label values period periodla
label def periodla 1 "1", modify
label def periodla 2 "2", modify
label def periodla 3 "3", modify
label def periodla 4 "4", modify
label def periodla 5 "5", modify
label def periodla 6 "6", modify
label def periodla 7 "7", modify

------------------

Whenever I do the regression like so

logit trans b1.period, or

I get the following

Iteration 0:   log likelihood =  -23971.55  
Iteration 1:   log likelihood = -23918.449  
Iteration 2:   log likelihood = -23918.271  
Iteration 3:   log likelihood = -23918.271  

Logistic regression                                     Number of obs = 55,847
                                                        LR chi2(5)    = 106.56
                                                        Prob > chi2   = 0.0000
Log likelihood = -23918.271                             Pseudo R2     = 0.0022

--------------------------------------------------------------------------------
         trans | Odds ratio   Std. err.      z    P>|z|     [95% conf. interval]
---------------+----------------------------------------------------------------
        period |
   2  |   .7467871   .0439266    -4.96   0.000     .6654699    .8380408
3  |   .8045088   .0275544    -6.35   0.000     .7522759    .8603684
  4  |   .7566667   .0595155    -3.55   0.000      .648565    .8827867
5  |    .854655     .05772    -2.33   0.020     .7486935     .975613
  6  |   .7256962   .0332995    -6.99   0.000     .6632792    .7939867
               |
         _cons |   .1974665   .0027929  -114.69   0.000     .1920677    .2030171

I put b1.period and get all of the other period categories (2,3,4,5,6) in my output table, APART from the final one , which is 7.
Any Ideas why 7 is not coming up on the regression output?

[William Lisowski]

One possibility is that logit also output a warning message — that you did not include in the output you shared — that warned you that every observation with period==7 had the same value of trans, so those observations were excluded because the coefficient cannot be estimated when there is no variability.

When you post output, it is more helpful to copy from the Stata Results window the command that was run and everything following it until the next command.

I might have been able to use your example data to check this guess, but you only showed values of period and not of your dependent variable.

[Joseph Coveney]

With nearly 56 000 observations distributed among seven categories of a variable named "period", I would not have guessed separation at first.

In lieu of dataex, which shows only a relative handful of observations, how about showing the result of

tabulate period trans, missing

or just attach the dataset.

[Josie Thompson]

Thank you both - all the period ==7 inputs had the same trans value aswell, so sorted it now!