A Monte Carlo study of alternative approaches for dealing with randomly missing data