Hi!
I know that results from regular -oaxaca- depend on the omitted base category for categorical variables, so it is recommended to use the normalize option.

Say for example, I have a predictor variable Religion that has 3 groups (religion2_1-religion2_3).
The detailed results will contain ALL religion categories, as opposed to regular regression with only 2 (k-1). How do I interpret the decomposition results for all the dummy variables?

. oaxaca contra_modern normalize(religion2_*), logit by(dhsyr) svy(,subpop(include_teen)) pooled swap detail
(normalized: religion2_1 religion2_2 religion2_3)

Blinder-Oaxaca decomposition

Number of strata = 116 Number of obs = 41,013
Number of PSUs = 2,024 Population size = 40,980.468
Design df = 1,908
Model = logit
Group 1: dhsyr = 1 N of obs 1 = 526
Group 2: dhsyr = 0 N of obs 2 = 379

------------------------------------------------------------------------------
| Linearized
contra_mod~n | Coef. Std. Err. t P>|t| [95% Conf. Interval]
-------------+----------------------------------------------------------------
overall |
group_1 | .2838129 .026565 10.68 0.000 .2317134 .3359124
group_2 | .2009951 .0210072 9.57 0.000 .1597956 .2421946
difference | .0828178 .033847 2.45 0.015 .0164367 .1491989
explained | -.0003794 .0023023 -0.16 0.869 -.0048948 .0041359
unexplained | .0831973 .0337456 2.47 0.014 .0170152 .1493794
-------------+----------------------------------------------------------------
explained |
religion2_1 | -.0003667 .0018479 -0.20 0.843 -.0039908 .0032573
religion2_2 | -.0000165 .0008027 -0.02 0.984 -.0015908 .0015579
religion2_3 | 3.76e-06 .000192 0.02 0.984 -.0003728 .0003803
-------------+----------------------------------------------------------------
unexplained |
religion2_1 | .1511666 .0581182 2.60 0.009 .0371848 .2651484
religion2_2 | -.006532 .0049029 -1.33 0.183 -.0161477 .0030837
religion2_3 | -.0052311 .011557 -0.45 0.651 -.0278968 .0174346
_cons | -.0562062 .0630383 -0.89 0.373 -.1798375 .067425