Ecosyste.ms: Awesome

An open API service indexing awesome lists of open source software.

Awesome Lists | Featured Topics | Projects

https://github.com/insightsengineering/unihtee

Tools for uncovering treatment effect modifiers in high-dimensional data.
https://github.com/insightsengineering/unihtee

heterogeneous-treatment-effects high-dimensional-data nonparametrics targeted-learning variable-importance

Last synced: 6 days ago
JSON representation

Tools for uncovering treatment effect modifiers in high-dimensional data.

Awesome Lists containing this project

README 

        
---

output: github_document

bibliography: inst/references.bib 

---



```{r, include = FALSE}

knitr::opts_chunk$set(

  collapse = TRUE,

  comment = "#>",

  fig.path = "man/figures/README-",

  out.width = "100%"

)

```



# R/`unihtee`



> Univariate Heterogeneous Treatment Effect Estimation



__Author:__ [Philippe Boileau](https://pboileau.ca/)



[![Project Status: Active – The project has reached a stable, usable state and is being actively developed.](https://www.repostatus.org/badges/latest/active.svg)](https://www.repostatus.org/#active)



---



`unihtee` provides tools for uncovering treatment effect modifiers in

high-dimensional data. Treatment effect modification is defined using variable

importance parameters based on absolute and relative effects. Inference is

performed about these variable importance measures using nonparametric

estimators. Users may use one-step or targeted maximum likelihood estimators.

Under general conditions, these estimators are unbiased and efficient.



Additional details about this methodology is provided in @boileau2022

and in the package's

[vignette](https://insightsengineering.github.io/unihtee/main/articles/using-unihtee.html).



## Installation



The package may be installed from GitHub using

[`remotes`](https://CRAN.R-project.org/package=remotes):



```{r gh-master-installation, eval=FALSE}

remotes::install_github("insightsengineering/unihtee")

```



`unihtee` is under active development. Check back often for updates.



## Usage



`unihtee()` is the only user-facing function. It can be used to perform

inference about the treatment effect modification variable importance

parameters. These parameters are defined for data-generating processes with

continuous, binary and time-to-event outcomes with binary exposure variables.

Variable importance parameters based on absolute and relative effects are

available. Details are provided in the vignette.



## Example



We simulate some observational study data that contains ten pre-treatment

covariates, of which are two treatment effect modifiers. We then perform

inference about the absolute treatment effect modifier variable importance

parameter, which is inspired by the average treatment effect.



```{r example}

library(unihtee)

library(MASS)

library(data.table)

library(sl3)



set.seed(510)



## create the dataset

n_obs <- 500

w <- mvrnorm(n = n_obs, mu = rep(0, 10), Sigma = diag(10))

confounder_names <- paste0("w_", seq_len(10))

colnames(w) <- confounder_names

a <- rbinom(n = n_obs, size = 1, prob = plogis(w[, 1] + w[, 2]))

y <- rnorm(n = n_obs, mean = w[, 1] + w[, 2] + a * w[, 3] - a * w[, 4])

dt <- as.data.table(cbind(w, a, y))



## targeted maximum likelihood estimates and testing procedure

unihtee(

  data = dt,

  confounders = confounder_names,

  modifiers = confounder_names,

  exposure = "a",

  outcome = "y",

  outcome_type = "continuous",

  effect = "absolute",

  estimator = "tmle"

)

```



## Issues



If you encounter any bugs or have any specific feature requests, please [file

an issue](https://github.com/insightsengineering/unihtee/issues).



## Contributions



Contributions are very welcome. Interested contributors should consult our

[contribution

guidelines](https://github.com/insightsengineering/unihtee/blob/master/.github/CONTRIBUTING.md)

prior to submitting a pull request.



## Citation



To cite `unihtee` and the papers introducing the underlying framework, use the

following BibTeX entries:



```

@manual{unihtee,

  title = {unihtee: Univariate Heterogeneous Treatment Effect Estimation},

  author = {Philippe Boileau},

  note = {R package version 0.0.1}

}



@misc{boileau2023,

      title={A nonparametric framework for treatment effect modifier discovery in high dimensions}, 

      author={Philippe Boileau and Ning Leng and Nima S. Hejazi and Mark van der Laan and Sandrine Dudoit},

      year={2023},

      eprint={2304.05323},

      archivePrefix={arXiv},

      primaryClass={stat.ME}

}



@article{boileau2022,

    author = {Boileau, Philippe and Qi, Nina Ting and van der Laan, Mark J and Dudoit, Sandrine and Leng, Ning},

    title = {A flexible approach for predictive biomarker discovery},

    journal = {Biostatistics},

    year = {2022},

    month = {07},

    issn = {1465-4644},

    doi = {10.1093/biostatistics/kxac029},

    url = {https://doi.org/10.1093/biostatistics/kxac029}

}

```



## License



The contents of this repository are distributed under the Apache 2.0 license.

See the

[`LICENSE.md`](https://github.com/insightsengineering/unihtee/blob/main/LICENSE.md)

and

[`LICENSE`](https://github.com/insightsengineering/unihtee/blob/main/LICENSE)

files for details.



## References