I am trying to simulate monotone and non-monotone missing data patterns on a repeated measures longitudinal dataset (with no missing values to begin with). For simplicity, lets assume 1000 observations (one record per subject), a group variable with 2 levels, with 4 repeated measures of an outcome variable.
https://blogs.sas.com/content/iml/2016/10/26/patterns-of-missing-data.html
I believe it's easy to simulate non-monotone missing data pattern using the example in the link above. And, in order to create monotone patterns, you can specify the patterns you need using a zero-one matrix and either use proc iml or data steps to get the end result.
I have the following questions -
1. It would be helpful if someone can share a worked example/code out there that creates both monotone and non-monotone missing data patterns, and perhaps allows you to control the amount/rate of missing data for these different patterns.
2. Also, can we use a propensity score based model to create monotone missing data pattern for a dataset?
Any links/articles/code would be helpful.
Thanks!