When losses are piecewise linear, optimal decision-making hinges on estimating a specified fractile of a probability distribution. In this article we investigate, through examples, the operating characteristics of an estimation procedure which utilizes simple linear combinations of one or two values in the order statistic generated by an observation vector. Such estimates appear to be remarkably efficient compared to the estimates generated by simple parametric procedures; moreover, this efficiency does not appear to be damaged excessively by the presence of moderate amounts of autocorrelation in the sample values. It also appears to be possible to exploit informal knowledge about the shape of the distribution to improve the efficiency of this procedure.
No takes yet. Share an insight, caveat, or question.
Robert H. Hayes (1971) studied this question.
Synapse has enriched 2 closely related papers on similar clinical questions. Consider them for comparative context: