Answers are grounded in this paper and cite their sources.
Two estimators of the median based on sequentially obtained quantal response data are studied. The design is essentially the up-and-down method and consists of taking K observations per trial at sequentially determined levels until some fixed total sample size is obtained. The asymptotic distribution of the sample average, , of the levels is studied and an estimate of its variance is proposed. The bias and mean square error of , which have been numerically evaluated for the normal distribution, indicate there is often little loss in efficiency using K up to 5 instead of the conventional K = 1. The asymptotic properties of the Spearman-Kärber estimator, , show that its large sample efficiency can decrease as the number of distinct levels increases. Small sample results using K = 1 indicate that is quite robust with respect to the choice of initial level and spacing between levels.
No takes yet. Share an insight, caveat, or question.
Robert K. Tsutakawa (1967) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: