Calculating the mean and variance of power sums with two log-normal components

An efficient method for accurately calculating the mean and variance of power sums with two log-normal components is presented. It involves numerical quadrature with integrands consisting of Gaussian probability density functions. Only the technique of transformation and the trapezoidal rule are used in the evaluation. The whole computational work and roundoff error are much lower than the one by Schwartz and Yeh (1982). Because of its high efficiency, no least-squares fit is necessary.<
>