A collaboration between Facilities Operations and the Department of Statistics at UConn on energy monitoring.

A hybrid (frequentist and Bayesian) and computationally intensive approach to anomaly detection, applied to natural gas usage data from UConn.

Bayesian network meta-regression hierarchical model with multivariate t-random effects, where the variances of the random effects are modeled as a log-linear function of related covariates.

Robustified Bayesian quantile regression models with sparse signal shrinkage prior to address model misspecification. Estimation is carried out with variational Bayes.