Bethany Wolf: Variable Importance Measures for Interpretable Machine Learning

American Statistical Association (ASA), Section on Statistical Learning and Data Science (SLDS) February webinar: Illuminating the Black Box: Variable Importance Measures for Interpretable Machine Learning Record: February 25, 2021 Presenter: Dr. Bethany Wolf is Professor of Biostatistics at the Medical University of South Carolina. She is currently the Associate Director of the Methodology Core for a program project grant with the Division of Rheumatology entitled “Improving Minority Health in Rheumatic Disease (I aM HeaRD)" and serves as the primary statistician for the Department of Anesthesia and Perioperative Medicine. Since becoming faculty at MUSC, she has served as a co-investigator of the Methodology Cores for an Multidisciplinary Clinical Research Center (MCRC) of Rheumatic Diseases in African Americans and the Biostatistics Core of the South Carolina Clinical and Translational Research Institute (SCTR). Her primary research is in the field of machine learning for developing statistical approaches to identify gene x gene and gene x environment interactions associated with patient outcomes without an a priori hypothesis and development of a new statistical hypothesis testing framework to evaluate whether it is appropriate to dichotomize a continuous predictor for disease discrimination. As a member of the CU and through her involvement in the Methodology Cores for the CTSA and the MCRC, she also has the opportunity to interact with scientist and clinical faculty across a multitude of disciplines at MUSC to design research studies and develop analysis plans for grant proposals, analyze a variety of data, and collaborate in writing manuscripts. Abstract: In this talk, Dr. Wolf will provide an overview of common machine learning methods (e.g. ensemble methods, support vector machines and neural networks) and discuss currently implemented importance measures that provide semi-quantitative measures of associations between variables and outcome. For more information about or to join ASA SLDS, visit https://community.amstat.org/slds/home https://www.amstat.org/