Darby Thompson

Darby JS Thompson, PhD

Statistics and Actuarial Science
Simon Fraser University
8888 University Drive
Burnaby, BC, Canada
V5A 1S6

Email: dthompso "AT" sfu.ca

Welcome to my webpage. Find a summary of some of my research interests, my CV, and the portal to materials for courses I may be teaching. Note that course material will go away when the semester is over, so if there's something you'd like to see and you can no longer access it, just send me an email.

Research Interests

Multi-Phase Hazard Models and Time-Dependent Covariate Effects in Time-to-Event Data

Darby Thompson I am interested in situations where time-to-event data cannot be adquately modeled with a proportional hazards model. In particular, when the hazard function may be composed of multiple phases. A motivating example is that of recovery from a harsh intervention where there is a short term period of increased risk due to the intervention (ie: surgery) itself before individuals begin to experience a treatment effect. Methods to address this problem include changepoint, mixture, and flexible spline methods which enable estimation of different effects of covariates over time. I am investigating and extending these methods and describing situations to effectively implement such models and interpret the parameters.

Joint Models for Longitudinal and Time-to-Event Data

I've recently become interested in utilizing longitudinal data to improve estimation in Time-to-Event models. Among other approaches, shared-parameter Joint Models can be implemented in this situation. These interesting and complex models assume that the longitudinal and time-to-event measurements are different representations of an underlying unmeasurable (latent) health status. In this fast-evolving field, there are many open problems including assessing treatment effects, efficiently estimating standard errors, diagnostics and model-selection. Despite these open problems, the utility of Joint Models to provide dynamic individual-level predictions and improve population differences appears superior to standard methods. Darby Thompson

Design and Analysis of Clinical Research Studies

Darby Thompson I am involved in supporting ongoing clinical studies as an advisor, consultant and principal statistician. Every study poses unique problems and I enjoy working with scientists in many diverse fields. Solving real problems for researchers is a strong motivation in all of my work.

Innovations in Statistical Education

Statistics classes don't have to be boring. I continually seek different methods to teach introductory and higher-level statistical courses through effective in-class examples/interaction and by utilizing multi-media in the classroom setting. Effective evaluation of student performance through assignment, project-work, and examinations are ongoing areas of interest for me.