Abstract
Two different designs, namely composites inside and composites over strata have been studied to be performed in a stratified population. The variance formula in composite sampling has been presented based on the variation between all the possible composites. According to this source of variation, we have proposed a new design, as a combination of composite and ranked set sampling improving the precision of the estimator to estimate the population mean. Finally, we have compared these designs to some traditional rival designs using two real case studies on medicinal flower and soil pollution.