CTSI to host webinar series on recent topics in research methods

UNIVERSITY PARK, Pa. — Penn State Clinical and Translational Science Institute announced its Biostatistics, Epidemiology and Research Design (BERD) Research Methods Seminar Series schedule for the spring semester. The series features lectures on fundamental research methods in the fall semester and covers more advanced topics in the spring semester. Statisticians and methodologists from multiple Penn State departments present these seminars.

Seminars are held virtually from 2 to 3 p.m. Connection information is sent upon registration.

The spring 2026 schedule is:

• Feb. 2: “Making Mediation Dynamic: Inference in Continuous-Time Mediation Models” by Ivan Jacob Pesigan
  • Mediation analysis helps researchers understand how and why interventions work, but traditional approaches depend heavily on how often data are collected, making results hard to interpret and compare across studies. This seminar introduces continuous-time mediation models, a more flexible approach that better captures how effects unfold over time, even when measurements are unevenly spaced. Using applied examples, Pesigan will show how this framework clarifies when mediation effects begin, how they change and which time periods matter statistically. The session will wrap up with practical guidance for researchers, including how to implement these methods using open-source cTMed R software and what they mean for study design in health and behavioral research.
  • Register here
• Feb. 16: “A Semi-Parametric Global Bandit Framework for Flexible and Scalable Sequential Decision-Making” by Hyebin Song
  • Sequential decision-making is at the heart of personalized medicine and adaptive clinical trials, where treatments must be adjusted as data accumulate. While multi-armed bandit methods are well suited to these problems, real-world data introduce complications like delayed feedback, patient characteristics, and related treatment options.
    
    In this talk, Song will present a new framework for batched bandit problems that accounts for these challenges. The approach uses a flexible yet interpretable model to learn how treatments are related and introduces a new algorithm that efficiently narrows down the best options over time. Through simulations and real-world examples, Song will show how this method improves decision-making accuracy and outperforms existing approaches in complex clinical settings.
  • Register here
• March 16: “Using Mixed Methods in Interventional Trials: An Overview” by Lauren Van Scoy
  • ​​​​​​​This presentation explores how mixed methods can strengthen interventional research by improving study design, implementation, and interpretation. Van Scoy will review common frameworks that guide when and how to integrate qualitative components into trials. Using real-world examples, the session will show how different mixed methods designs help clarify the purpose and value of qualitative data, including its use as an outcome measure. Attendees will leave with practical strategies for adding rigor while preserving real-world context.
  • Register here
• April 7: “McCartanIdentification and Semiparametric Estimation of Conditional Means from Aggregate Data” by Cory McCartan
  • ​​​​​​​This webinar introduces a new method for estimating group-level outcomes when only summarized data, such as regional averages, is available. McCarten will discuss how common approaches rely on overly strong assumptions and will present a more flexible, machine learning–based alternative. Using simulations and real-world examples, McCarten will show how this method improves accuracy, allows researchers to test key assumptions, and provides reliable estimates even with complex data. Open-source tools will be shared so attendees can apply the approach in their own work.
  • Register here
• April 20: “Forecasting and Diagnosing Stress-Induced Physical Activity Decline with Deep Learning” by Young Won Cho
  • ​​​​​​​Major disruptions, like injury or the COVID-19 lockdown, can cause sudden drops in physical activity, but people recover in very different ways. Identifying who is likely to stay disengaged early on is critical for timely intervention, yet difficult when activity patterns are noisy and highly individualized. This webinar introduces a new, two-step approach that combines modern machine learning with interpretable statistical methods to detect stress-related declines in physical activity. Using wearable device data from an aging study during the COVID-19 lockdown, we show how pre-disruption activity patterns can be used to forecast future behavior and identify distinct recovery profiles. Through real-world examples, Cho will highlight how advanced models can improve early detection while still producing results that are practical, transparent, and actionable for health researchers and intervention designers.
  • Register here

BERD Core is part of Penn State Clinical and Translational Science Institute. To request a consultation with the core, fill out a service request form.