In this workshop, Dr. Schnitzer gives an overview of the estimation of causal effects using Targeted Learning and how this approach can be applied in drug safety and effectiveness research.
Targeted Learning (TL) focuses on the clear delineation of assumptions, definition of the parameter of interest, usage of high-performance statistical methods to target estimation of the parameter of interest, and careful interpretation of the resulting estimate. The estimation step involves the usage of Targeted Minimum Loss-based Estimation (van der Laan and Rubin, 2006) and an ensemble learning method known as Super Learner (van der Laan, Pulley and Hubbard, 2007).
This talk will qualitatively contrast traditional statistical modeling (e.g. model fitting) and flexible prediction methods (e.g. machine learning) with the TL approach. Special focus will be placed on estimation in high-dimensional covariate spaces and variable selection.
Analyst session: Following the lecture on Targeted Learning by Mireille Schnitzer, Menglan Pang provides concrete examples of how to apply TMLE in a study.
After learning the theories and assumptions of targeted learning (see lecture by Mireille Schnitzer), this presentation focuses on the technical implementation of targeted maximum likelihood estimation (TMLE). This talk is intended for analysts, but will give all viewers a more concrete understanding of the TMLE procedure.
This talk by CNODES analyst Menglan Pang covers the following sections:
• Step-by-step description of TMLE procedure
• Implementation of TMLE using SAS
• Double robustness of TMLE: A simulation study
• Application of TMLE on a real study