Over the summer, one of our newest faculty members, Dr. Jon Steingrimsson, gave an invited talk, entitled “Censoring Unbiased Survival Forests” at the 2017 Lifetime and Data Analysis Conference hosted by the University of Connecticut. At the conference, Steingrimsson was awarded “Best Paper” for his work entitled “Censoring Unbiased Regression Trees and Ensembles”.
Steingrimsson’s research which he began while at Johns Hopkins and accomplished with colleagues L. Diao from the University of Waterloo and R. Stawderman from the University of Rochester, proposes a novel approach to building regression trees and ensemble learning in survival analysis. Semiparametric theory for missing data is used to construct two specific classes of methods for building regression trees and regression ensembles that respectively make use of Buckley-James and doubly robust estimating equations for a given full data risk function. The research further shows how to implement these algorithms using existing software for uncensored responses.