computing frequency distribution from data