ABSTRACT
An experiment was done with the objects of estimating: the difference in mean strength of two worsted yarns produced by slightly different processes, and the variation of strength between and within bobbins for yarns of this type. The numbering of bobbins within each yarn is random, as too is the numbering of test lengths within bobbins. The components of variance have at least three uses. Firstly, they provide summary descriptions of important aspects of the problem under study and so may be regarded as primary parameters. Next, they clarify the fact that to compare the means of the two yarns, the mean square between bobbins within yarns provides an appropriate estimate of error. Finally, via a process of synthesis of variance, it is possible to estimate the consequences of some different scheme of sampling in which the number of repeat observations per bobbin is changed.
