The Expected Value of Perfect Information in Facility Location

This paper is concerned with finding the expected value of perfect information (EVPI) in a simple facility location problem where the weights, which summarize cost and volume parameters, are random draws from a multivariate probability distribution. A model with rectangular distances and one with squared Euclidean (center of gravity) are used. The analysis is developed for a multivariate normal distribution of weights but simulation is used to show that the expressions derived are reasonably accurate for other distributions (the EVPI for the multivariate lognormal distribution is illustrated).