Concepedia

Abstract

These remarks should not be construed to mean that cluster sampling should always be avoided. In many research situations it will be found to be the most efficient design, despite the effect of the intraclass correlation on the variance. This effect, however, must be accounted for by using the proper formulas in computing the variance.9 When this is done, probability statements about the results can legitimately be formulated in the standard manner. 9 For the computational formula for the variance of the mean where the sample has been drawn by cluster sampling procedures see: ibid., p. 420; similarly for the variance of a proportion see: Philip J. McCarthy, Sample Design, in Marie Jahoda, Morton Deutsch, and Stuart W. Cook, (editors), Research Methods in Social Relations, Vol. 11, New York: Dryden Press, 1951, p. 662.