In this article, we propose two-layer median ranked set sampling (TMRSS) design that combines median ranked set sampling (MRSS) and two-layer ranked set sampling (TRSS). Ranked set sampling (RSS) is an alternative sampling method that can improve the efficiency of estimators when exact measurement of response variable is either difficult, time consuming or expensive. Evaluation of the TMRSS performance for different distributions, set, and cycle sizes regarding mean and regression coefficients estimators and mean square of the regression model are carried out using Monte Carlo simulation study and real data application. The results indicate that estimators of TMRSS yields are either equivalent to or better than MRSS.