Lectures and seminars Faculty event: Patrick Schnell
Speaker: Patrick Schnell, Ohio State University College of Public Health
Title
Topics in Selection Bias and Causal Inference with Counting Processes
Abstract
I will give an overview of recent, current, and upcoming research projects to help identify collaboration opportunities. As part of its COVID-19 monitoring and control strategy, The Ohio State University required students living or attending classes on-campus to be tested once per week, with those testing positive subsequently isolated. The proportion of tests returning positive results was used as a daily estimate of prevalence. We showed that this test-positive rate is a biased estimate of prevalence under this and other seemingly innocuous longitudinal testing schemes due to the structure of the repeated testing schedule, even in the absence of other confounding factors: time since last negative test affects both probability of testing and infectiousness. We proposed an unbiased estimator based on Horvitz-Thompson inverse probability of sampling weighting.
I will also present current work on graphical tools for control of selection bias based on potential outcomes. These tools can be used to combine multiple samples, selected via differing mechanisms, to identify causal effects. I will give examples of identifiability arguments for case-cohort studies, multiple/time-dependent exposures, and causal effects of selection itself.
Finally, I will introduce the starts of two projects using ideas from causal analysis of competing risks: a standardization approach to comparing the occurrence of adverse events between treatments when treatment affects survival, and a potential competing events formulation of multi-state models for treatment switching.
About the speaker
I am faculty in the Division of Biostatistics in The Ohio State University College of Public Health, specializing in the applied areas of clinical trials, electronic medical records, and cancer survivorship. My methodological research focuses on Bayesian hierarchical models and computation, causal inference, and more recently selection bias. I am planning a sabbatical visit to KI for the 2025-2026 academic year to learn more about epidemiologic methods and practice, especially using national registers and surveys.
Zoom link
https://ki-se.zoom.us/j/64279974590?pwd=L2lCeHJlckt1dy9aUzN6RTJkNFlHQT09
Meeting ID: 642 7997 4590
Passcode: 437012
Join the faculty event in Wargentin! After the seminar, the faculty lunch will be served in Ljusgården.