Double percentile ranked set sampling procedure (DPRSS) is introduced as a modification of ranked set sampling (RSS) for estimating the mean of the population of interest. The efficiency of the DPRSS estimator of the mean relative to the simple random sampling (SRS), ranked set sampling (RSS), median ranked set sampling (MRSS), extreme ranked set sampling (ERSS) and percentile ranked set sampling (PRSS) methods is obtained. It turns out that DPRSS is an unbiased and more efficient than its counterpart SRS, RSS, MRSS, ERSS, and PRSS for the symmetric distributions. For non-symmetric distributions considered in this study, the DPRSS estimator has a smaller bias and it is more efficient than the SRS, ERSS and PRSS methods.