Scholion.ai

A Note on Mean Square Successive Differences

J. N. K. Rao

Journal of the American Statistical Association · 1959 · Volume 54 · Issue 288

DOI: 10.1080/01621459.1959.11683600

Abstract

This paper deals with the use of higher order mean square successive differences for estimating dispersion when a strong trend is present in the mean value and the method of first differences does not adequately eliminate the trend. The first four moments of the mean square successive second difference, δ 2 2 , are derived in samples from an arbitrary population with constant mean (or with linear trend in the means). Because the efficiency of δ 2 2 relative to δ 2 2 increases with β 2 , the approximate distribution of δ 2 2 is discussed only for a leptokurtic population, specifically for the symmetrical two-tailed exponential population. For populations with varying mean the first two moments of δ 2 2 are given and the effect of trend of the mean values on these moments is discussed. Short tables of the efficiency of the mean square successive second, third, and fourth differences, all relative to s 2 are given.

Enable JavaScript for citation tools and additional paper interactions.