From: Comparison of methods for handling covariate missingness in propensity score estimation with a binary exposure
Group
x5
I{Y > mean(Y)}
% of observations
% missing on x1
1
0
25
10
20
2
40
3
30
60
4
80
Total
50