Concomitant-based rank set sampling proportion estimates.

This paper discusses the rank set sampling (RSS) protocol as it pertains to the estimation of a population proportion. The ranking process is based on a concomitant variable. The concomitant-based RSS estimate is asymptotically normal so standard inference procedures can still be implemented. This is illustrated using a real data set from the medical literature. The performance of the estimator is studied in terms of relative efficiency. Generally speaking, the concomitant-based RSS estimator is more efficient than the proportion of successes in a simple random sample. The greatest gains in efficiency are obtained when the correlation between the Bernoulli and concomitant variable is large in absolute value.