How to generate Missing Not at random (MNAR) data in R?

Question

I have Y being binomially distributed with probability 0.1 and N=100 . And independent variable x with normally distributed mean 0 and 0.5. and I want to generate MNAR mechanism on Y.

Answer 1

To generate MNAR data you can use a data generating process where the missingness mechanism depends on the unobserved data.

# Generate the true data
y1 <- rbinom(100, size=1, prob=0.1)
# Generate the missing process. Depends on the "true" observed value
r  <- rbinom(length(y1), size=1, prob=c(.25, .1)[y1+1])
y  <- y1
y[r==1] <- NA