When I select a sample using > *Restrict, based in criterion *and then > *Use dummy variable* all works fine, and I get the appropriate sub-set. When I then select > *Random sub-sample* from this subset, everything is fine as long as the number selected is less than the number of items that were selected on the first call. However, there is a misleading message *Number of observations to select (max x)*, where x is greater than the number of observations that were selected on the first call. Furthermore, if I specify a number larger than the number of obs that were selected, it executes the specified x. However, on occasions I do get the message *Invalid number of cases*. If I re-enter the same x then it goes through. It appears there's a problem in trapping this. Thanks. *"In God we trust, all others bring data."*