A Belief Function Approach to Likelihood Updating in a Gaussian Linear Model
P M Meehan, A. P. Dempster, Emery N. Brown · 1996
Abstract We consider a stochastic model for cortisol dynamics that can be expressed as a Gaussian linear model conditional on a set of unknown pulse locations. Implementation of the Bayesian paradigm by Monte Carlo Markov Chajn (MCMC) methods requires iterative updating of the likelihood for changes in the model. We show that likelihood updating can be efficiently performed by representing the model components as a set of belief functions and using the sweep operator to perform computations.