← Back to Event List

Stat Colloquium [In-Person]: Dr. Joyee Ghosh

The University of Iowa


Mathematics/Psychology : 401

Date & Time

October 13, 2023, 11:00 am12:00 pm


Title: Prediction of North Atlantic Tropical Cyclone Activity via Bayesian Model Averaging


Seasonal forecasting of the frequency of North Atlantic tropical storms is of interest, because it can provide basic information towards improved preparation against these storms. It has been shown that sea surface temperatures (SSTs) during the hurricane season can predict tropical cyclone (TC) activity well. But predictions need to be made before the beginning of the hurricane season, when the predictors are not yet observed. Several climate models issue forecasts of the SSTs, which can be used instead. Such models use the forecasts of SSTs as surrogates for the true (yet to be observed) SSTs. In the first half of the talk, we develop a fully Bayesian negative binomial regression model, which makes a distinction between the true SSTs and their forecasts, both of which are included in the model. For prediction, the true SSTs may be regarded as missing/unobserved predictors and sampled from their posterior predictive distribution. Our model can simultaneously handle missing predictors and variable selection uncertainty. If the main goal is prediction, an interesting question is: should we include predictors in the model that are missing at the time of prediction? 

In the second half of the talk our focus is on prediction of multiple metrics that measure the activity, intensity, severity etc of a hurricane season. Average sea surface temperatures (SSTs) during the hurricane season have been used as predictors for each of these metrics, in separate univariate regression models, in the literature. When the main objective is prediction of the response variables, a natural question is: do multivariate regression models that accommodate dependency among the response variables improve prediction compared to their univariate counterparts? Whether Bayesian multivariate normal regression models improve prediction compared to their univariate counterparts is not clear from the existing literature, and in this work, we try to fill this gap.