Targeted Maximum Likelihood Estimation for Network Data

Estimation of average causal effects for single time point interventions in network-dependent data (e.g., in the presence of spillover and/or interference). Supports arbitrary interventions (static or stochastic). Implemented estimation algorithms are the targeted maximum likelihood estimation (TMLE), the inverse-probability-of-treatment (IPTW) estimator and the parametric G-computation formula estimator. Asymptotically correct influence-curve-based confidence intervals are constructed for the TMLE and IPTW. The data are assumed to consist of rows of unit-specific observations, each row i represented by variables (F.i,W.i,A.i,Y.i), where F.i is a vector of friend IDs of unit i (i's network), W.i is a vector of i's baseline covariates, A.i is i's exposure (can be binary, categorical or continuous) and Y.i is i's binary outcome. Exposure A.i depends on (multivariate) user-specified baseline summary measure(s) sW.i, where sW.i is any function of i's baseline covariates W.i and the baseline covariates of i's friends in F.i. Outcome Y.i depends on sW.i and (multivariate) user-specified summary measure(s) sA.i, where sA.i is any function of i's baseline covariates and exposure (W.i,A.i) and the baseline covariates and exposures of i's friends. The summary measures are defined with functions def.sW and def.sA. See ?'tmlenet-package' for a general overview.


[![Travis-CI Build Status](https://travis-ci.org/osofr/tmlenet.svg?branch=master)](https://travis-ci.org/osofr/tmlenet) [![Coverage Status](https://coveralls.io/repos/osofr/tmlenet/badge.png?branch=master&service=github)](https://coveralls.io/r/osofr/tmlenet?branch=master)

The tmlenet R package performs estimation of average causal effects for single time point interventions in network-dependent (non-IID) data in the presence of interference and/or spillover. Currently implemented estimation algorithms are the targeted maximum likelihood estimation (TMLE), Horvitz-Thompson or the inverse-probability-of-treatment (IPTW) estimator and the parametric G-computation estimator. The user-specified interventions can be either static, dynamic or stochastic. Asymptotically correct influence-curve-based confidence intervals are also constructed for the TMLE and IPTW. See the paper below for more information on the estimation methodology employed by the tmlenet R package:

To install the development version of tmlenet (requires the devtools package):

devtools::install_github('osofr/tmlenet', build_vignettes = FALSE)

Once the package is installed, please refer to the help file ?'tmlenet-package' and tmlenet function documentation for details and examples:

?'tmlenet-package'
?tmlenet

The input data are assumed to consist of rows of unit-specific observations, with each row i represented by variables (F.i,W.i,A.i,Y.i), where F.i is a vector of "friend IDs" of unit i (also referred to as i's "network"), W.i is a vector of i's baseline covariates, A.i is i's exposure (either binary, categorical or continuous) and Y.i is i's binary outcome.

Each exposure A.i depends on (possibly multivariate) baseline summary measure(s) sW.i, where sW.i can be any user-specified function of i's baseline covariates W.i and the baseline covariates of i's friends in F.i (all W.j such that j is in F.i). Similarly, each outcome Y.i depends on sW.i and (possibly multivariate) summary measure(s) sA.i, where sA.i can be any user-specified function of i's baseline covariates and exposure (W.i,A.i) and the baseline covariates and exposures of i's friends (all W.j,A.j such that j is in i's friend set F.i).

The summary measures (sW.i,sA.i) are defined simultaneously for all i with functions def.sW and def.sA. It is assumed that (sW.i,sA.i) have the same dimensionality across i. The function eval.summaries can be used for evaluating these summary measures.

All estimation is performed by calling the tmlenet function. The vector of friends F.i can be specified either as a single column in the input data (where each F.i is a string of friend IDs or friend row numbers delimited by character sep) or as a separate input matrix of network IDs (where each row is a vector of friend IDs or friend row numbers). Specifying the network as a matrix generally results in significant improvements to run time. See tmlenet function help file for additional details on how to specify these and the rest of the input arguments.

...

To cite tmlenet in publications, please use:

Sofrygin O, van der Laan MJ (2015). tmlenet: Targeted Maximum Likelihood Estimation for Networks. R package version 0.1.

The development of this package was funded through an NIH grant (R01 AI074345-07).

This software is distributed under the GPL-2 license.

News

tmlenet 0.0.9

  • Initial release version

Reference manual

It appears you don't have a PDF plugin for this browser. You can click here to download the reference manual.

install.packages("tmlenet")

0.1.0 by Oleg Sofrygin, 2 years ago


https://github.com/osofr/tmlenet


Report a bug at https://github.com/osofr/tmlenet/issues


Browse source code at https://github.com/cran/tmlenet


Authors: Oleg Sofrygin [aut, cre], Mark J. van der Laan [aut]


Documentation:   PDF Manual  


GPL-2 license


Imports assertthat, data.table, Matrix, methods, R6, Rcpp, simcausal, speedglm, stats, stringr

Suggests doParallel, foreach, igraph, knitr, locfit, matrixStats, RUnit

Linking to Rcpp


Suggested by simcausal.


See at CRAN