Statistical Distributions in Univariate and Multivariate Edgeworth Populations
Andrew W. Davis · Biometrika · 1976
A general method is presented for the formal construction of statistical distributions in univariate and multivariate populations, when the underlying distributions are specifiable by Edgeworth-type expansions. It is shown that the distribution of a statistic may be derived in principle from the appropriate normal-theory noncentral distribution, by means of a symbolic expectation operator. This provides a unified approach to the results of numerous investigations into the effects of nonnormality on various well-known statistics.