Schedule

Regular meeting time: Every other Friday 1:00 - 2:00 pm, Skylab Conference Room (UOP 3-351) or online  (with possible different day/time/place for webinars/invited talks)

September 25, Friday 

9:00 - 10:00 AM in the Fishbowl Conference Room (UOP 254): watch an ISBS webinar together. Title: "Sensitivity analyses for estimating treatment effects in post-baseline biomarker responders in cardiovascular outcome studies" by Drs. Arianna Nuti from University of Florence and Björn Bornkamp from Novartis AG. Abstract: Post-baseline biomarker responder analyses in clinical trials face well-known methodological challenges, including selection bias from observed and unobserved patient characteristics as well as immortal-time bias in time-to-event outcomes. This presentation reviews this problem and proposes a sensitivity analysis framework for estimating treatment effects in the principal stratum of the biomarker responders on treatment using a Gaussian copula approach. The methodology transparently separates quantities that can be identified from observed data (the treatment-specific distributions of biomarker and clinical outcomes) from those requiring unverifiable assumptions about cross-treatment dependencies. Rather than imposing simple conditional independence assumptions, we parameterize residual dependence between potential outcomes across treatment arms through a copula, which is then varied in sensitivity analyses. We present the approach through a motivating example based on cardiovascular trials, develop the framework and illustrate it on a simulated data set.

1:00-2:00 PM in the Skylab Conference Room (UOP 351): (1) introduction and welcome to the first meeting of the semester (snacks will be provided!) and (2) discuss the webinar from the morning. 

October 9, Friday

1:00 - 2:00 PM in Skylab (UOP 351): Dr. Eaton will present about an in-progress analysis for a collaborative project. There is an interesting time-to-event element because this data is cross sectional, but the people invited to do the study are 5-20 years out from surgery. Also, not everyone agreed to participate in the study, and the people getting the surgery has been changing over time during the period from 2005-2020, so it is interesting to use statistical methods to answer what would outcomes have looked like if the whole sample got surgery and then were followed for 5-20 years?

October 23, Friday

1:00 - 2:00 PM in Skylab (UOP 351): We will read Dr. Bin Nan's paper, "Kernel estimation of bivariate time-varying coefficient model for longitudinal data with terminal event" (JASA, 2024, 119:1102-1111) at our working group meeting (a student volunteer is needed to lead the discussion). Dr. Nan will give a seminar, entitled "Modeling Longitudinal Data with Terminal Event" in our division on 10/21/2026.