From: Comparison of methods for handling covariate missingness in propensity score estimation with a binary exposure
Group
x8
I{Y > mean(Y)}
% of observations
% missing on x3
1
0
40
30
60
2
25
50
3
10
15
4
20
Total