Short course: Causal Inference with Structural Nested Models

Judith J. Lok, PhD, associate professor of Mathematics and Statistics, Boston University, and adjunt associate professor of biostatistics, Harvard T.H. Chan School of Public Health, will discuss the G-estimation of Structural Nested Models (SNMs), a method designed to estimate the causal effect of a time-varying treatment in the presence of treatment-confounder feedback. The course will describe g-estimation of SNMs both for continuous outcomes (Structural Nested Mean Models) and time-to-event outcomes (Structural Nested Failure Time Models).

The course will take place November 22, 9:30am-4:30pm, TMEC 250. Registration required.

Abstract
When a time-varying treatment is repeatedly adapted to evolving prognostic factors and the value of the prognostic factors is affected by prior treatment, the effect of the treatment cannot simply be estimated by conditioning on the patient characteristics. This treatment-confounder feedback is common in observational studies. G-estimation of Structural Nested Models is a method designed to estimate the causal effect of a time-varying treatment in the presence of treatment-confounder feedback. The course will describe g-estimation of Structural Nested Models (SNMs), both for continuous outcomes (Structural Nested Mean Models or SNMMs) and time-to-event outcomes (Structural Nested Failure Time Models or SNFTMs). I will describe methods with and without rank preservation, using single-robust and doubly robust estimators. I will present methods for estimating and testing the parameters of SNMMs. I will also compare SNMs with Marginal Structural Models (MSMs). I will illustrate all methods with examples from HIV/AIDS.

Prerequisites
Basic knowledge of SAS or R (I will work with SAS but programming can also be done with R). Basic concepts in biostatistics, including familiarity with regression and logistic regression.

A wi-fi enabled laptop with SAS or R installed.

Judith J. Lok, PhD
Judith Lok is an associate professor of mathematics and statistics at Boston University, and an adjunct associate professor of biostatistics at the Harvard T.H. Chan School of Public Health. Her research focuses on Causal Inference and Survival Analysis. Research topics in Causal Inference include Structural Nested Models, Marginal Structural Models, Inverse Probability of Censoring Weighting (both for MAR and MNAR data), and mediation analysis. Lok also works on adaptive clinical trial designs, in particular Learn-As-you-GO or LAGO. Her main application area is HIV/AIDS; other application areas are bacterial infections, HCV, and maternal and child health.

Download course materials

Course Slides [PDF]
Course Lab Information [PDF]
Course Lab Simulated Data Set in CSV format [CSV]

Course Lab Simulated Data Set in SAS Format [SAS]