Abstract:
As a modification of ranked set sampling (RSS), multistage median RSS
(MMRSS), is used for distribution function estimation. The performance of the
empirical distribution function obtained by MMRSS is examined in terms of relative efficiency. The effect of the set size, number of cycles, and number of stages
of MMRSS on the performance of the proposed estimator is addressed. Generally
speaking, the empirical distribution function when using MMRSS is more efficient
than when using simple random sampling for some quantiles. The comparison is
carried out theoretically and numerically.