roboBayes

An implementation of Bayesian online changepoint detection (Adams and MacKay (2007) ) with an option for probability based outlier detection and removal (Wendelberger et. al. (2021) ). Building on the independent multivariate constant mean model implemented in the 'R' package 'ocp', this package models multivariate data as multivariate normal about a linear trend, defined by user input covariates, with an unstructured error covariance. Changepoints are identified based on a probability threshold for windows of points.