Clinical Research Statistical Methods Seminars

The Clinical Research Statistical Methods Seminars will be held monthly or bimonthly, depending on interest (Location: San Francisco VA Medical Center). Our goal is to provide an opportunity for those based at the SFVAMC (statisticians, epidemiologists, data analysts, etc.) to discuss methodology related to clinical research studies. We want to provide a friendly, informal setting in which trainees and junior colleagues have an opportunity to present their work, seek feedback on methodological approaches, and learn about recent developments in statistics. Please contact Rebecca Scherzer or Craig Meyer for more information.


Upcoming seminars:


Date: Monday, June 12th, 1PM – 2PM

Presented by: Craig S. Meyer, PhD

Location: Director’s Conference Room – 3rd floor, Building 210

Topic:  Causal inference in the presence of time-varying confounding: an application of the parametric g-formula

Abstract:  Causal inference from longitudinal observational studies is limited in the presence of time-varying confounding. A class of causal inference models called G-methods have been developed to adjust effect estimates and improve the causal interpretation of those estimates. This talk will present one of these methods known as the parametric g-formula, a flexible model that allows the investigator to simulate counterfactual scenarios. Recent applications have included investigations of HIV treatment, hypothetical interventions in the treatment of diabetes and coronary artery disease, among others. A discussion of identifying time-varying confounding and the parametric g-formula will be presented. An example will follow applying the g-formula to an occupational cohort investigating occupational exposures to ionizing radiation and the risk of cataracts in medical radiologic technologists.

Past Seminars:


Monday, May 8th, 2017: Rebecca Scherzer, Ph.D.

Topic: A tutorial on accounting for competing risks in survival analysis

Abstract: Clinical research studies often record the time to more than one outcome, such as death or end stage renal disease (ESRD). Standard survival analysis methods such as Cox proportional hazards regression are not appropriate to analyze such data, where an individual who experiences a competing event (such as death) is treated as censored for the outcome of interest (ESRD). In this seminar, we will discuss alternatives such as the Fine-Gray competing-risks analysis, which extends the Cox model to competing-risks data by simultaneously evaluating hazards for the primary (ESRD) and competing (death) events. We will also review current recommendations and newer methods for calculating cumulative incidence, hazard ratios, and number needed to treat, as discussed in the recent paper by Austin and Fine (Statistics in Medicine, 2017).


March 2017:  Martin Frigaard:  K nearest neighbors and the curse of dimensionality

Sept 2016: Rebecca Scherzer: structural equation modeling of blood pressure and kidney function

July 2016: Debbie Huang: impact of lung transplantation on health-related quality of life

June 2016: Craig Meyer: use of CART and other boosted regression tree methods

March 2016: Anna Rubinsky: analytic approaches and pitfalls in modeling BP and eGFR trajectories

January 2016: Ruth Dubin: vascular dysfunction in ESRD

October 2015: Rebecca Scherzer: analytic techniques for longitudinal biomarker data