Kristin Linn, PhD
Assistant Professor
Department of Biostatistics
University of Pennsylvania

“Confounding in Imaging-based Predictive Modeling”

The multivariate pattern analysis (MVPA) of                neuroimaging data typically consists of one or more statistical learning models applied within a broader image analysis pipeline.  The goal of MVPA is often to learn about patterns of variation encoded in magnetic resonance images (MRI) of the brain that are associated with brain disease incidence, progression, and response to therapy.  Every model choice that is made during   image processing and analysis can have implications with respect to the results of neuroimaging studies.  Here, attention is given to two important steps within the MVPA framework: 1) the standardization of features prior to training a supervised learning model, and 2) the training of learning models in the presence of confounding.  Specific examples focus on the use of the support vector machine, as it is a common model choice for MVPA, but the general concepts apply to a large set of models employed in the field.  We propose novel methods that lead to improved classifier performance and interpretability, and we illustrate the methods on real neuroimaging data from a study of Alzheimer’s disease.

José R. Zubizaretta, PHD
Associate Professor, Department of Health Care Policy, Harvard Medical School
Faculty Affiliate, Department of Statistics , Faculty of Arts and Sciences
Harvard University

“Building Representative Matched Samples with Multi-valued Treatments in Large Observational Studies”

 In observational studies of causal effects, matching methods are widely used to approximate the ideal study that would be conducted under controlled experimentation. In this talk, I will discuss new matching methods that use tools from modern optimization to overcome five limitations of standard matching approaches. In particular, these new matching methods (i) directly obtain flexible forms of covariate balance, as specified before matching by the investigator; (ii) produce self-weighting matched samples that are representative of target populations by design; and (iii) handle multiple treatment doses without resorting to a generalization of the propensity score. (iv) These methods can handle large data sets quickly. (v) Unlike standard matching approaches, with these new matching methods, usual estimators are root-n consistent under usual conditions. I will illustrate the performance of these methods in a case study about the impact of a natural disaster on educational opportunity.