interpreting a logistic regression with interaction terms

[Elisa Peset]

I am examining how the association between pregnancy intent (0=wanted; 1=unwanted) and birth weight (0=low birth weight; 1=non-low birth weight) differs across intersectional strata (different combinations of maternal age, education, and residency type. A variable was made with the categories being all possible combinations of these three variables. These are the variable categories:
Maternal age: 0 = <25; 1= 25 or higher
Education: 0 = up to primary; 1 = secondary or higher
Residency type: 0 = rural 1 = urban

The categories for the intersectional variable are:
0 = <25/up to primary/rural
1 = <25/up to primary/urban
2 = <25/secondary or higher/rural
3 = <25/secondary or higher/urban
4 = ≥25/up to primary/rural
5 = ≥25/up to primary/urban
6 = ≥25/secondary or higher/rural
7 = ≥25/secondary or higher/urban



Could someone help me interpret the output? I am not sure how to interpret the interaction terms and the strata (non-interaction)/



. svy: logistic birthweight i.pregintent_bin##ib6.strata if pop==1
(running logistic on estimation sample)

Survey: Logistic regression

Number of strata = 1 Number of obs = 7,931
Number of PSUs = 7,931 Population size = 7,325.9934
Design df = 7,930
F(15, 7916) = 2.85
Prob > F = 0.0002

---------------------------------------------------------------------------------------------------------
| Linearized
birthweight | Odds ratio std. err. t P>|t| [95% conf. interval]
----------------------------------------+----------------------------------------------------------------
pregintent_bin |
unwanted | .7117476 .1909799 -1.27 0.205 .4206223 1.20437
|
strata |
<25 up to pimary rural | .7811451 .4002248 -0.48 0.630 .2861189 2.132637
<25 up to pimary urban | 2.734558 .7791837 3.53 0.000 1.564261 4.78041
<25 secondary or higher rural | .6428229 .1970044 -1.44 0.149 .3525217 1.172187
<25 secondary or higher urban | 1.602041 .4983672 1.51 0.130 .8706411 2.947869
24> up to primary rural | 1.634665 .7085737 1.13 0.257 .6988827 3.823429
24> up to primary urban | 1.159998 .3437606 0.50 0.617 .6488864 2.0737
24> secondary or higher urban | 1.08384 .4377024 0.20 0.842 .491089 2.392051
|
pregintent_bin#strata |
unwanted#<25 up to pimary rural | 2.543342 1.713579 1.39 0.166 .6789217 9.527742
unwanted#<25 up to pimary urban | .9174659 .398509 -0.20 0.843 .3915678 2.149675
unwanted#<25 secondary or higher rural | 2.455064 1.002542 2.20 0.028 1.102594 5.4

[Stephen Jenkins]

Welcome to Statalist. Please take a few minutes to read the Forum FAQ. (Click on the link at the top of page within line with black background.) Note well how you should post Stata output using CODE delimiters, and the instructions how to do this. It's easy and improves legibility and re-usability hugely.

Substantive advice: (a) read appropriate texts about using and interpreting interaction effects in logit models. (Note that odds ratio interpretations are deprecated by many in this context. (See e.g. Norton and Dowd, "Log Odds and the Interpretation of Logit Models", Health Services Research, DOI: 10.1111/1475-6773.12712) (b) For a sophisticated command to handle interactions, look at Rendean's -ginteff- in the Stata Journal (https://journals.sagepub.com/doi/pdf...6867X231175253)