A0167
Title: Normalized latent measure models
Authors: Jim Griffin - University College London (United Kingdom) [presenting]
Abstract: Normalized latent measure models (NLMMs) are a framework for modeling and comparing probability distributions using mixtures of nonparametric distributions. Although the methods are generic, the focus is on modeling similar distributions. For example, to model a large collection of probability distributions (such as areal income distributions) or the effects of covariates on a probability distribution (the effect of wind direction and speed on energy generation). The framework allows understanding the heterogeneity in the distributions, and the variation is attributed to spatial factors or other covariates. As well as introducing the models, it is discussed how identifiability, variable selection, and overfitting can be addressed within a Bayesian framework to provide interpretable inferential methods. Their use is illustrated in a range of applications.
