Moments of the sum of correlated log-normal random variables

Abstract

This paper presents a comparison of Fenton's and Schwartz and Yeh's methods concerning their capability of predicting the mean and the variance of the sum of a finite number of correlated log-normal random variables. The predictions by the two methods differ more with decreased correlation between the log-normal components, with increasing number of components and increasing component variance. Schwartz and Yeh's method seems to provide more realistic predictions for the mean and the standard deviation of the power sum by showing a closer agreement with the results of an exact formulation for the outage probability in mobile radio systems.