Causal Decomposition of Group Disparities

Estimates the causal decompositions of group disparities developed by Yu and Elwert (2025) . For the nuisance functions of the estimators, we provide both parametric and nonparametric options, as well as manual options in case the default models are not satisfying.

cdgd

The package cdgd implements the causal decompositions of group disparities in Yu and Elwert (2025).

Installation

The latest release of the package can be installed through CRAN.

install.packages("cdgd")

The current development version can be installed from source using devtools.

devtools::install_github("ang-yu/cdgd")

Examples

library(cdgd)  

# load the simulated example data
data(exp_data)
head(exp_data)
#>       outcome treatment  confounder          Q group_a
#> 748 1.4608165         1  0.26306864  0.6748330       0
#> 221 0.4777308         0  1.30296394  0.5920512       1
#> 24  0.8760129         1 -1.49971226  1.6294327       1
#> 497 0.4131192         1 -1.17219619 -0.8391873       1
#> 249 2.0483222         1  1.71790879  2.9546966       1
#> 547 0.1912013         0 -0.02438458 -0.3704544       0

Use cdgd0_ml, cdgd0_pa, or cdgd0_manual for unconditional decomposition

results0 <- cdgd0_pa(Y="outcome",D="treatment",G="group_a",X=c("confounder","Q"),data=exp_data,alpha=0.05)

round(results0$results, 4)
#>              point     se p_value CI_lower CI_upper
#> total       0.2675 0.0390  0.0000   0.1911   0.3439
#> baseline    0.0421 0.0131  0.0013   0.0164   0.0678
#> prevalence  0.2579 0.0337  0.0000   0.1919   0.3240
#> effect     -0.1372 0.0209  0.0000  -0.1781  -0.0963
#> selection   0.1047 0.0150  0.0000   0.0754   0.1340

Use cdgd1_ml, cdgd1_pa, or cdgd1_manual for conditional decomposition

results1 <- cdgd1_pa(Y="outcome",D="treatment",G="group_a",X="confounder",Q="Q",data=exp_data,alpha=0.05)

round(results1, 4)
#>                                 point     se p_value CI_lower CI_upper
#> total                          0.2675 0.0390  0.0000   0.1911   0.3439
#> baseline                       0.0421 0.0131  0.0013   0.0164   0.0678
#> conditional prevalence         0.2032 0.0371  0.0000   0.1305   0.2760
#> conditional effect            -0.1644 0.0220  0.0000  -0.2076  -0.1212
#> conditional selection          0.0875 0.0143  0.0000   0.0595   0.1156
#> Q distribution                 0.0990 0.0188  0.0000   0.0621   0.1359
#> conditional Jackson reduction  0.2362 0.0378  0.0000   0.1621   0.3103