In tlverse/tmle3mopttx: Targeted Maximum Likelihood Estimation of the Mean under Optimal Individualized Treatment

knitr::opts_chunk$set(
  collapse = TRUE,
  comment = "#>",
  fig.path = "README-"
)

R/tmle3mopttx

Targeted Learning and Variable Importance with Optimal Individualized Categorical Treatment

Authors: Ivana Malenica, Jeremy R. Coyle and Mark J. van der Laan

Description

Suppose one wishes to maximize (or minimize) the population mean of an outcome where for each individual one has access to measured baseline covariates. We consider estimation of the mean outcome under the optimal rule, where the candidate rules are restricted to depend only on user-supplied subset of the baseline covariates (or no covariates specified). The estimation problem is addressed in a statistical model for the data distribution that is nonparametric.

The tmle3mopttx is an adapter/extension R package in the tlverse ecosystem, that estimates the mean outcome under the following regimes:

(1) Optimal Individualized Treatment for categorical treatment, (2) Optimal Individualized Treatment based on possibly sub-optimal rules (including static rules), (3) Optimal Individualized Treatment based on realistic rules. (4) Optimal Individualized Treatment with resource constraints.

The tmle3mopttx also provides variable importance analysis in terms of optimizing the population mean of an outcome for settings (1)-(3). It also can return an interpretable rule based on a Highly Adaptive Lasso fit.

In addition, tmle3mopttx supports estimating the mean outcome under regimes under the following missigness process:

1. Missing outcome.

In order to avoid nested cross-validation, tmle3mopttx relies on split-specific estimates of the conditional expectation of the outcome ($\mathbb{E}_0(Y|A,W)$) and conditional probability of treatment ($P_0(A|W)$) in order to estimate the rule. The targeted maximum likelihood estimates of the mean performance under the estimated rule are obtained using CV-TMLE.

The description of the method implemented, with simulations and data examples using tmle3mopttx can be found in @malenica2021opttx. For additional background on Targeted Learning and previous work on optimal individualized treatment regimes, please consider consulting @vdl2007super, @cvtmle2010, @vdl2011targeted, @vanderLaanLuedtke15, @luedtke2016super, @jeremythesis and @vdl2018targeted.

---

Installation

You can install the most recent stable release from GitHub via devtools with:

devtools::install_github("tlverse/tmle3mopttx")

---

Citation

After using the tmle3mopttx R package, please cite the following:

@software{malenica2022tmle3mopttx,
      author = {Malenica, Ivana and Coyle, Jeremy and {van der Laan}, Mark J},
      title = {{tmle3mopttx}: Targeted Learning and Variable Importance with Optimal Individualized Categorical Treatment},
      year  = {2022},
      doi = {},
      url = {https://github.com/tlverse/tmle3mopttx},
      note = {R package version 1.0.0}
    }

Issues

If you encounter any bugs or have any specific feature requests, please file an issue.

---

License

The contents of this repository are distributed under the GPL-3 license. See file LICENSE for details.

---

References

tlverse/tmle3mopttx documentation built on Aug. 9, 2022, 3:31 p.m.