Abstract
When sample sizes vary, analysis by control charts must be modified from that for equal sample sizes. A simple weighted average of the 's, with weights n, is taken for in general. Usually when n's are small, ranges R are used. Then to estimate σ for the process (if variabilities are indeed homogeneous) an unweighted average of the unbiased estimates, R/d2, has commonly been used. This paper provides two systems of simple weights from which better results are obtainable. Similarly, weights for sample standard deviations are also given.

This publication has 0 references indexed in Scilit: