Daniel Almirall, of the Data Science for Dynamic Intervention Decision-Making Lab in the Institute for Social Research at the University of Michigan, will present:

“Longitudinal Mixed-effects Modeling to Compare Dynamic Treatment Regimens with SMART Data”

Abstract: A dynamic treatment regimen (DTR) is a pre-specified sequence of decision rules which map baseline or time-varying measurements on an individual to a recommended intervention or set of interventions. Sequential multiple assignment randomized trials (SMARTs) represent one type of data collection tool for informing the construction of DTRs. Most SMARTs have multiple DTRs embedded within the trial. A common primary aim in a SMART is the marginal mean comparison between two (or more) of the embedded DTRs. This manuscript develops a longitudinal mixed-effect modeling and estimation approach for marginal mean comparisons of the embedded DTRs in a SMART with a continuous, repeated measures outcome. This work includes developing an unbiased estimator for marginal mean comparison(s) of embedded DTRs (primary aim), as well as a proposal for estimating the random effects (which is often of secondary interest in the analysis of randomized trials). The methodology is illustrated using data from a SMART used to develop a DTR for improving social communication in children with autism spectrum disorder.

This talk is based on joint work with Brook Luers, Inbal Nahum-Shani, Connie Kasari and Min Qian.

A social tea will be held at 9:30 a.m. in A434 Mayo. All are Welcome.