haldensify

haldensify: Highly adaptive lasso conditional density estimation in R - Published in JOSS (2022)

https://github.com/nhejazi/haldensify

Keywords

Keywords from Contributors

Scientific Fields

Repository

:package: R/haldensify: Highly Adaptive Lasso Conditional Density Estimation

README.Rmd

---
output:
  rmarkdown::github_document
always_allow_html: yes
bibliography: "inst/REFERENCES.bib"
---



```{r, echo = FALSE}
knitr::opts_chunk$set(
  collapse = TRUE,
  comment = "#>",
  out.width = "100%",
  fig.path = "README-"
)
```


# R/`haldensify`


[![R-CMD-check](https://github.com/nhejazi/haldensify/workflows/R-CMD-check/badge.svg)](https://github.com/nhejazi/haldensify/actions)
[![Coverage Status](https://img.shields.io/codecov/c/github/nhejazi/haldensify/master.svg)](https://app.codecov.io/github/nhejazi/haldensify?branch=master)
[![CRAN](http://www.r-pkg.org/badges/version/haldensify)](https://www.r-pkg.org/pkg/haldensify)
[![CRAN downloads](https://cranlogs.r-pkg.org/badges/haldensify)](https://CRAN.R-project.org/package=haldensify)
[![CRAN total downloads](http://cranlogs.r-pkg.org/badges/grand-total/haldensify)](https://CRAN.R-project.org/package=haldensify)
[![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)
[![MIT license](http://img.shields.io/badge/license-MIT-brightgreen.svg)](https://opensource.org/licenses/MIT)
[![DOI](https://zenodo.org/badge/DOI/10.5281/zenodo.3698329.svg)](https://doi.org/10.5281/zenodo.3698329)
[![DOI](https://joss.theoj.org/papers/10.21105/joss.04522/status.svg)](https://doi.org/10.21105/joss.04522)


> Highly Adaptive Lasso Conditional Density Estimation

__Authors:__ [Nima Hejazi](https://nimahejazi.org), [David
Benkeser](https://sph.emory.edu/profile/faculty/david-benkeser), and Mark
van der Laan](https://vanderlaan-lab.org/about/)

---

## What's `haldensify`?

The `haldensify` R package is designed to provide facilities for nonparametric
conditional density estimation based on a flexible procedure proposed initially
by @diaz2011super. The core of the implemented methodology involves recovering
conditional density estimates by performing pooled hazards regressions so as to
assess the conditional hazard that an observed value falls in a given bin over
the (conditional) support of the variable of interest. Such conditional density
estimates are useful, for example, in causal inference problems in which the
_generalized propensity score_ (for continuous-valued exposures) must be
estimated [@diaz2012population; @diaz2018stochastic; @diaz2020causal].
`haldensify` implements this conditional density estimation strategy for use
only with the highly adaptive lasso (HAL) [@benkeser2016highly;
@vdl2017generally; @vdl2018highly; @coyle2022hal9001-rpkg;
@hejazi2020hal9001-joss]. Since the generalized propensity score is a key
ingredient in inverse probability weighting (IPW) methods, `haldensify` builds
on the advances of @ertefaie2020nonparametric and @hejazi2022efficient to
provide non-parametric IPW estimators of the causal effects for continuous
treatments, which achieve the semi-parametric efficiency bound by
undersmoothing along a family of HAL conditional density estimators.

---

## Installation

For standard use, we recommend installing the package from
[CRAN](https://CRAN.R-project.org/package=haldensify) via

```{r cran-installation, eval = FALSE}
install.packages("haldensify")
```

To contribute, install the _development version_ of `haldensify` from GitHub
via [`remotes`](https://CRAN.R-project.org/package=remotes):

```{r gh-master-installation, eval = FALSE}
remotes::install_github("nhejazi/haldensify")
```

---

## Example

A simple example illustrates how `haldensify` may be used to train a highly
adaptive lasso model to obtain conditional density estimates:

```{r example-fit}
library(haldensify)
set.seed(76924)

# simulate data: W ~ U[-4, 4] and A|W ~ N(mu = W, sd = 0.25)
n_train <- 100
w <- runif(n_train, -4, 4)
a <- rnorm(n_train, w, 0.25)

# HAL-based density estimate of A|W
haldensify_fit <- haldensify(
  A = a, W = w,
  n_bins = 10, grid_type = "equal_range",
  lambda_seq = exp(seq(-1, -10, length = 100)),
  # arguments passed to hal9001::fit_hal()
  max_degree = 3,
  reduce_basis = 1 / sqrt(n_train)
)
haldensify_fit
```

We can also visualize the empirical risk (with respect to density loss) in terms
of the solution path of the lasso regularization parameter:
```{r example-plot, out.width = "80%", fig.path = "man/figures/", fig.alt = "CV-risk of regularized conditional density estimators"}
plot(haldensify_fit)
```

Finally, we can obtain conditional density estimates from the trained model on
the training (or on new) data:
```{r example-predict}
# use the built-in predict method to get predictions
pred_haldensify <- predict(haldensify_fit, new_A = a, new_W = w)
head(pred_haldensify)
```

For more details, check out the [package
vignette](https://codex.nimahejazi.org/haldensify/articles/intro_haldensify) on
the corresponding `pkgdown` site.

---

## Issues

If you encounter any bugs or have any specific feature requests, please [file
an issue](https://github.com/nhejazi/haldensify/issues).

---

## Contributions

Contributions are very welcome. Interested contributors should consult our
[contribution
guidelines](https://github.com/nhejazi/haldensify/blob/master/CONTRIBUTING.md)
prior to submitting a pull request.

---

## Citation

After using the `haldensify` R package, please cite the following:

        @article{hejazi2022efficient,
          author = {Hejazi, Nima S and Benkeser, David and D{\'\i}az, Iv{\'a}n
            and {van der Laan}, Mark J},
          title = {Efficient estimation of modified treatment policy effects
            based on the generalized propensity score},
          year = {2022},
          journal = {},
          publisher = {},
          volume = {},
          number = {},
          pages = {},
          doi = {},
          url = {https://arxiv.org/abs/2205.05777}
        }

        @article{hejazi2022haldensify-joss,
          author = {Hejazi, Nima S and {van der Laan}, Mark J and Benkeser,
            David C},
          title = {{haldensify}: Highly adaptive lasso conditional density
            estimation in {R}},
          year  = {2022},
          doi = {10.21105/joss.04522},
          url = {https://doi.org/10.21105/joss.04522},
          journal = {Journal of Open Source Software},
          publisher = {The Open Journal}
        }

        @software{hejazi2022haldensify-rpkg,
          author = {Hejazi, Nima S and Benkeser, David C and {van der Laan},
            Mark J},
          title = {{haldensify}: Highly adaptive lasso conditional density
            estimation},
          year  = {2022},
          howpublished = {\url{https://github.com/nhejazi/haldensify}},
          doi = {10.5281/zenodo.3698329},
          url = {https://doi.org/10.5281/zenodo.3698329},
          note = {{R} package version 0.2.5}
        }

---

## Related

* [R/`hal9001`](https://github.com/tlverse/hal9001) -- The highly adaptive
  lasso estimator used internally to constructed conditional density estimates.

---

## Funding

The development of this software was supported in part through grants from the
National Library of Medicine (award number [T32
LM012417](https://reporter.nih.gov/project-details/9248418)), the
National Institute of Allergy and Infectious Diseases (award number
[R01 AI074345](https://reporter.nih.gov/project-details/9926564)) of the
National Institutes of Health, and the National Science Foundation (award
number [DMS 2102840](https://www.nsf.gov/awardsearch/showAward?AWD_ID=2102840)).

---

## License

© 2019-2025 [Nima S. Hejazi](https://nimahejazi.org)

The contents of this repository are distributed under the MIT license. See below
for details:
```
MIT License

Copyright (c) 2019-2025 Nima S. Hejazi

Permission is hereby granted, free of charge, to any person obtaining a copy
of this software and associated documentation files (the "Software"), to deal
in the Software without restriction, including without limitation the rights
to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
copies of the Software, and to permit persons to whom the Software is
furnished to do so, subject to the following conditions:

The above copyright notice and this permission notice shall be included in all
copies or substantial portions of the Software.

THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
SOFTWARE.
```

---

## References

Owner

JOSS Publication

haldensify: Highly adaptive lasso conditional density estimation in R

Published

September 23, 2022

DOI

10.21105/joss.04522

Volume 7, Issue 77, Page 4522

Authors

Nima S. Hejazi
Department of Biostatistics, T.H. Chan School of Public Health, Harvard University

Mark J. van der Laan
Division of Biostatistics, School of Public Health, University of California, Berkeley, Department of Statistics, University of California, Berkeley

David Benkeser
Department of Biostatistics and Bioinformatics, Rollins School of Public Health, Emory University

Editor

Mark A. Jensen

Tags

machine learning causal inference conditional density estimation generalized propensity score inverse probability weighting semiparametric inference

View PDF Review Thread Software Archive

GitHub Events

Committers

Last synced: 7 months ago

Issues and Pull Requests

Last synced: 6 months ago