Simulating numeric data according to distribution assumptions