How to graph results of logistic regression in Stata?

[Monzur Alam]

Dear all,

I am trying to examine the relationship between education and a woman’s probability of getting married, using a discrete time logistic regression model. The dependent variable is married (=1 or 0). For controls, I have a categorical variable for the individual’s own level of education, edu_cat (where 0 is no education, 1 and 2 are primary and secondary schooling respectively), and a bunch of other predictors. My Stata code and results are below.

. logistic married pyear pyearsq i.edu_cat i.bict muslim father_5andaboveeduc mother_5andaboveeduc hh_lowinc i.rural_div [pw= popweight]
 
Logistic regression                             Number of obs     =     30,336
                                                Wald chi2(16)     =    3150.08
                                                Prob > chi2       =     0.0000
Log pseudolikelihood = -1.856e+08               Pseudo R2         =     0.1996
 
--------------------------------------------------------------------------------------
                     |               Robust
             married | Odds Ratio   Std. Err.      z    P>|z|     [95% Conf. Interval]
---------------------+----------------------------------------------------------------
               pyear |   5.214708   .3749294    22.97   0.000     4.529288    6.003853
             pyearsq |   .9621039   .0020803   -17.87   0.000     .9580353    .9661898
                     |
             edu_cat |
                  1  |   .9944425   .0526095    -0.11   0.916     .8964955    1.103091
                  2  |   .7840981   .0457952    -4.16   0.000      .699288    .8791941
                     |
                bict |
birthcohort 1976-86  |   1.072139   .0569778     1.31   0.190     .9660838    1.189837
birthcohort 1987-97  |   1.036028   .0625042     0.59   0.557     .9204877    1.166071
                     |
              muslim |   1.299013   .0886869     3.83   0.000     1.136318    1.485003
father_5andaboveeduc |   .8964348   .0458894    -2.14   0.033     .8108581    .9910432
mother_5andaboveeduc |   .8650096   .0561416    -2.23   0.025     .7616852    .9823502
           hh_lowinc |   1.054659   .0639445     0.88   0.380     .9364904    1.187738
                     |
           rural_div |
            Barisal  |   1.185689   .1008533     2.00   0.045     1.003619     1.40079
         Chittagong  |    1.04448   .0626355     0.73   0.468     .9286565    1.174749
             Khulna  |   1.447291   .1105905     4.84   0.000     1.245988    1.681117
           Rajshahi  |   1.298267    .090639     3.74   0.000     1.132236    1.488644
            Rangpur  |   1.429192   .1046735     4.88   0.000     1.238081    1.649804
             Sylhet  |   .7301493   .0478659    -4.80   0.000     .6421108    .8302585
                     |
               _cons |   1.47e-08   8.71e-09   -30.37   0.000     4.58e-09    4.70e-08
--------------------------------------------------------------------------------------

I am presenting my analysis at a meeting, and would like to make my results more visually engaging, thereby presenting graphs rather than a bunch of regression tables. I was just wondering what command to use if I want to graph the relationship between the outcome variable, married, (which is binary) and the dependent variable, edu_cat, which is categorical.

Thank you!

Monzur

[Marcos Almeida]

I suggest you use the - margins - followed by the - marginsplot - command.

[Monzur Alam]

It worked, thank you very much!

Best,

Monzur

[Jeph Herrin]

I like to present regression model results as forest plots. See for example https://www.ctspedia.org/do/view/CTS...ClinAEGraph001 (click to enlarge). It's a way to visualize the effects and their precision, as well as their direction and significance, all at once. For logit models, this can be ORs, with a vertical line at 1.0.

hth,
Jeph
'