Tmle: AnRPackage for Targeted Maximum Likelihood Estimation

Susan Gruber,Mark J Van Der Laan

doi:10.18637/jss.v051.i13

Abstract

Targeted maximum likelihood estimation (TMLE) is a general approach for constructing an efficient double-robust semi-parametric substitution estimator of a causal effect parameter or statistical association measure. tmle is a recently developed R package that implements TMLE of the effect of a binary treatment at a single point in time on an outcome of interest, controlling for user supplied covariates, including an additive treatment effect, relative risk, odds ratio, and the controlled direct effect of a binary treatment controlling for a binary intermediate variable on the pathway from treatment to the out- come. Estimation of the parameters of a marginal structural model is also available. The package allows outcome data with missingness, and experimental units that contribute repeated records of the point-treatment data structure, thereby allowing the analysis of longitudinal data structures. Relevant factors of the likelihood may be modeled or fit data-adaptively according to user specifications, or passed in from an external estimation procedure. Effect estimates, variances, p values, and 95% confidence intervals are provided by the software.

Highlights

Research in fields such as econometrics, biomedical research, and epidemiology can involve collecting data on a sample from a population in order to assess the population or group level effect of a treatment, exposure, or intervention on a measurable outcome of interest
The default bound for g is set to (0.025, 0.975), but that guideline is flexible, and the effect on the bias and variance of the estimate depends on the data, e.g. if all values fall between (0.025,0.975), setting bounds closer to (0,1) will have no effect at all
As illustrated by the plots in the figure and the results reported in Table 2, depending http://biostats.bepress.com/ucbbiostat/paper275

Summary

Introduction

Research in fields such as econometrics, biomedical research, and epidemiology can involve collecting data on a sample from a population in order to assess the population or group level effect of a treatment, exposure, or intervention on a measurable outcome of interest. Obtaining an unbiased and efficient estimate of the statistical parameter of interest necessitates accounting for potential bias introduced through model misspecification, informative treatment assignment, or missingness in the outcome data. Due to the curse of dimensionality, parametric estimation approaches are not feasible for high dimensional data without restrictive simplifying modeling assumptions. Targeted maximum likelihood estimation (TMLE) is an efficient, double robust, semi-parametric methodology that has been successfully applied in these settings (van der Laan and Rubin 2006; van der Laan, Rose, and Gruber 2009). The development of the tmle package for the R statistical programming environment (Team 2011) was motivated by the growing need for a user-friendly tool for effective semiparametric estimation. The development of the tmle package for the R statistical programming environment (Team 2011) was motivated by the growing need for a user-friendly tool for effective semiparametric estimation. tmle is available for download from the Comprehensive R Archive Network at http://cran.r-project.org/web/packages/tmle/

Methods

Results

Conclusion