A unified ranked set sampling scheme is proposed to estimate the population mean. In such a scheme various existing ranked set sampling schemes are combined in order to minimizing the error of ranking and cost of sampling. It is shown that the sample weighted mean of the proposed scheme is more efficient than simple random sample mean. Also, assuming the underlying distribution is normal, the existence and uniqueness of maximum likelihood estimator of the location parameter are investigated. The pairwise relative precisions of the derived estimators are compared using simulation and numerical computations. It is concluded that a combination of existing sampling schemes may be considered as a good suggestion with rather high efficiency. A cost analysis is also performed. Some conclusions are eventually stated.

Keywords