With such a general question, I can only offer general ideas.
-- generate the complete data for the simulation study.
-- use a data step and a random number generator (ranuni, say) to identify data points to make missing.
This has the disadvantage of only generating missing data randomly. That means that analyses will work just like you have complete data with an effectively lower sample size.
If you want to make the missing data dependent on other variables, you can just nest the missing generation in IF statements. That dependency is the other extreme of missing data analysis.
You may wish to read the literature on analyzing missing data, as that can help in the strategy for creating missing data to meet your analytic goals. There are probably some references in the PROC MI chapter of the SAS/Stat manual.
