Implementing double-robust estimators of causal effects
Richard Emsley
Biostatistics, Health Methodology Research Group
The University of Manchester, UK
[email protected]
|
Mark Lunt
Arthritis Research Campaign Epidemiology Unit
The University of Manchester, UK
|
Andrew Pickles
Biostatistics, Health Methodology Research Group
The University of Manchester, UK
|
Graham Dunn
Biostatistics, Health Methodology Research Group
The University of Manchester, UK
|
Abstract.
This article describes the implementation of a double-robust estimator for
pretest–posttest studies (Lunceford and Davidian, 2004, Statistics in
Medicine 23: 2937–2960) and presents a new Stata command (dr)
that carries out the procedure. A double-robust estimator gives the analyst
two opportunities for obtaining unbiased inference when adjusting for
selection effects such as confounding by allowing for different forms of model
misspecification; a double-robust estimator also can offer increased
efficiency when all the models are correctly specified. We demonstrate the
results with a Monte Carlo simulation study, and we show how to implement the
double-robust estimator on a single simulated dataset, both manually and by
using the dr command.
View all articles by these authors:
Richard Emsley, Mark Lunt, Andrew Pickles, Graham Dunn
View all articles with these keywords:
dr, double-robust estimators, causal models, confounding, inverse probability of treatment weights, propensity score
Download citation: BibTeX RIS
Download citation and abstract: BibTeX RIS
|
