From: Comparison of methods for handling covariate missingness in propensity score estimation with a binary exposure
Group
x9
I{Y > mean(Y)}
% of observations
% missing on x4
1
0
20
15
30
2
25
50
3
35
70
4
40
Total