Computational Precision Health Seminar
Adaptive Sequential Design for a Single Time-Series
Date: Thursday, February 24, 2022
Time: 12:00 – 1:30 pm Pacific
Access: Join Zoom meeting, Meeting ID: 918 9154 9888 / Passcode: 596649
Speaker: BIDS Data Science Fellow Ivana Malenica, researcher at the Center for Targeted Machine Learning and Causal Inference, and PhD candidate in the Biostatistics Division at UC Berkeley
Abstract: Advances in digital technology have made it possible to deliver health interventions to individuals in their natural environment. In a micro-randomized trial, each individual is repeatedly randomized among multiple intervention options, often hundreds or even thousands of times over the course of the trial. In this talk, I motivate the need for robust statistical methods for precision health delivered via digital technology. Specifically, I will focus on a sequential adaptive design for a single individual which adapts the randomization mechanism for future time-point experiments. The aim of the design is to learn the optimal, unknown choice of the controlled components through time, while optimizing the expected outcome with inference. More generally, I discuss my contributions to the field of statistical precision health by (1) proposing several target parameters in a micro-randomized trial, including a general class of averages of conditional causal parameters defined by the current context; (2) motivating different exploration-exploitation strategies and approaches to learning the optimal individualized treatment (OIT); (3) studying the data-adaptive inference of the mean under the OIT, where the target adapts over time in response to the observed context of the individual.
CPH Seminars are hosted by the joint UC Berkeley-UCSF Program in Computational Precision Health.
Speaker(s)
Ivana Malenica
Ivana Malenica is a Ph.D. student in the Biostatistics Division working with Mark van der Laan, Antoine Chambaz and Alan Hubbard. She earned her Master’s in Biostatistics and Bachelor’s in Mathematics, and spent a year working as a Freeport-McMoRan research fellow in Data Science and Bioinformatics at the Translational Genomics Research Institute (TGen). Some of her prior work centers around mathematical modeling and Bayesian models for allele specific expression. Very broadly, her research interests span non/semi-parametric theory, probability theory, machine learning, causal inference and high-dimensional statistics. Most of her current work involves complex dependent settings (dependence through time and network) and adaptive sequential designs. She is also interested in model selection criteria, optimal individualized treatment, sensitivity analysis, mediation, online learning and software development (ex: medltmle, tstmle, tstmle01, sl3, cvma, tmle3opttx). Malenica is also one of the founding members of the tlverse software ecosystem, and works as a biostatistician on multiple projects at the Kaiser Permanente Research Division, TGen and the Bill & Melinda Gates Foundation.