Biostatistics Short Course: Statistical Methods for Truncated Time-to-Event Data: Concepts, Methods, and Applications

Please note: Registration coming soon. 

This short course will provide an overview of statistical methods for analyzing time-to-event data subject to truncation. Truncation arises when an event time is observed only if it falls within a specified observation window, resulting in a biased sample from the population of interest. Truncated time-to-event data occur in a wide range of applications, including epidemiologic cohort studies, studies of disease onset and survival, registry-based research, and other biomedical and public health settings. Depending on the sampling mechanism, data may be subject to left truncation, right truncation, or more complex forms of sequential truncation.

This course, led by Jing Qian, PhD, University of Massachusetts Amherst, will review the concept of truncation and highlight its distinction from censoring. Discussion will focus on classical methods for analyzing left-truncated time-to-event data, with or without additional right censoring, including risk-set adjustment methods for estimating event-time distributions and fitting regression models. Methods for right-truncated data will also be covered. Particular attention will be given to the assumption of quasi-independence between truncation and event times, which underlies many classical approaches, as well as hypothesis tests for assessing this assumption. The short course will then cover more recent developments for settings in which classical assumptions or standard sampling structures may not apply. Topics will include estimation and regression under dependent truncation, sequential truncation arising in observational cohort studies with complex sampling schemes, and methods for estimation and regression under sequential truncation. The methods will be illustrated using data examples, and R software will be used to demonstrate their implementation.

Instructor:

Jing Qian, PhD, Professor of Biostatistics, University of Massachusetts Amherst

Audience:

The course is intended primarily for biostatisticians, epidemiologists, quantitative researchers, and other investigators who work with time-to-event data and would like to better understand the methodological and practical issues that arise when such data are subject to truncation. Participants should have a basic understanding of probability and statistical inference, along with familiarity with standard survival analysis concepts, including survival and hazard functions, censoring, the Kaplan–Meier estimator, and Cox proportional hazards regression. Prior knowledge of statistical methods for truncated data is not required.

Participants are encouraged to bring a laptop for R demonstrations and optional short hands-on exercises. However, the course is not primarily a computing workshop, and the main content can be followed without running the code. Example code and datasets will be provided.