On some methods of analysis for weather experiments

Abstract

Using generalized densities, maximum likelihood and ancillarity, methods are developed for the analysis of randomized weather modification experiments that incorporate predictor variables. The linear logistic form is used to model the distributional singularity at the origin, while a gamma model with predictors is developed for the component conditional on the presence of precipitation. Connexions between the component submodels are explored and illustrated using an Australian data set.