Loading...
Thumbnail Image
Publication

Estimation of Causal Effects in Complex Clustered Data

Abstract
Analysis of clustered data from randomized trials or observational data often poses theoretical and practical statistical challenges, including but not limited to small numbers of independent units, many adjustment variables, continuous exposures, and/or differential clustering across trial arms. Further, commonly-used parametric methods rely on assumptions that may be violated in practice. Motivated by three scientific questions in public health, methods are developed and/or demonstrated for non-parametric estimation of causal effects. In Chapter 1, methods are elaborated for a cluster randomized trial (CRT) with missing individual-level data at baseline and follow-up, a complex sampling strategy, and limited number of clusters. Chapter 2 extends methods for CRTs with few independent units and partial clustering by trial arm. Chapter 3 demonstrates the theory and practical implementation of causal effect estimation for a ‘shift’ of a continuous exposure.
Type
openaccess
article
dissertation
Date
Publisher
Rights
License
http://creativecommons.org/licenses/by/4.0/
Research Projects
Organizational Units
Journal Issue
Embargo
Publisher Version
Embedded videos
Collections