UUÖ±²¥

Alexandra M. Schmidt, PhD
Professor of Statistics – Department of Statistical Methods, Mathematics Institute,
Federal University of Rio de Janeiro, Brazil

Hierarchical dynamic models for time series of highly structured biomedical signals

ALL ARE WELCOME

Abstract:

This talk considers the modelling of a set of temporally correlated signals with multilevel structure. The study is motivated by Raynaud’s phenomenon (RP) which results from blood vessel spasm. Individuals in the study belong to one of three groups, two groups showing different forms of RP and a control group. For each individual multiple time series were generated through thermal infrared imaging. These signals represent the temperature of 10 fingertips across time and document the thermal recovery of an individual from a standardized cold stress. To investigate the recovery temperature profiles of individuals from different groups, we propose a hierarchical dynamic model with an exponential growth component. For each group, the model provides estimates of the mean temperatures across time and of the increase in mean temperatures after the cold stress. These estimates allow us to learn how the temperature recovery differs among groups and how correlated the temperatures among fingers are in different groups. Inference is performed under the Bayesian paradigm, which naturally accounts for uncertainty about parameter estimates. Though motivated by the RP study, the proposed methodology is generally applicable to more general hierarchical time-varying outcomes and can accommodate different temporal structures.

Bio:

Alexandra M. Schmidt is a Professor of Statistics of the Federal University of Rio de Janeiro (UFRJ), Brazil. Her main areas of research are spatial statistics, spatio-temporal modeling, dynamic linear models, and hierarchical models. Her projects in these areas usually follow the Bayesian paradigm to perform inference. She completed her doctoral training at the University of Sheffield in 2001, under the supervision of Professor Anthony O’Hagan, worked for one year as a post-doctoral fellow with Professor Alan Gelfand at the University of Connecticut, and has been working at the Department of Statistical Methods, UFRJ, since 2002. She has published 29 papers in peer-reviewed journals, including JRSS Series B, JRSS Series C, Annals of Applied Statistics, Canadian Journal of Statistics, Technometrics, Environmetrics, and Test, and has supervised or co-supervised 9 Ph.D. and 14 M.Sc. students. In 2008 she was awarded with the Abdel El-Shaarawi Young Investigator Award of The International Environmetrics Society (TIES). She is the current President of the International Society for Bayesian Analysis (ISBA), and serves as Associate Editor of the Scandinavian Journal of Statistics (2012-), Environmetrics (2010-), and Stat (2011-). Please visit for more details.