nanxstats/msaenet

🧲 Multi-step adaptive estimation for reducing false positive selection in sparse regressions

msaenet

msaenet implements the multi-step adaptive elastic-net (MSAENet) algorithm for feature selection in high-dimensional regressions proposed in Xiao and Xu (2015) [PDF].

Nonconvex multi-step adaptive estimations based on MCP-net or SCAD-net are also supported.

Check vignette("msaenet") to get started.

Installation

You can install msaenet from CRAN:

install.packages("msaenet")

Or try the development version on GitHub:

remotes::install_github("nanxstats/msaenet")

Citation

To cite the msaenet package in publications, please use

Nan Xiao and Qing-Song Xu. (2015). Multi-step adaptive elastic-net: reducing false positives in high-dimensional variable selection. Journal of Statistical Computation and Simulation 85(18), 3755–3765.

A BibTeX entry for LaTeX users is

@article{xiao2015multi,
  title   = {Multi-step adaptive elastic-net: reducing false positives in high-dimensional variable selection},
  author  = {Nan Xiao and Qing-Song Xu},
  journal = {Journal of Statistical Computation and Simulation},
  volume  = {85},
  number  = {18},
  pages   = {3755--3765},
  year    = {2015},
  doi     = {10.1080/00949655.2015.1016944}
}

Gallery

Adaptive Elastic-Net / Multi-Step Adaptive Elastic-Net

Adaptive MCP-Net / Multi-Step Adaptive MCP-Net

Adaptive SCAD-Net / Multi-Step Adaptive SCAD-Net

Contribute

To contribute to this project, please take a look at the Contributing Guidelines first. Please note that the msaenet project is released with a Contributor Code of Conduct. By contributing to this project, you agree to abide by its terms.

License

msaenet is free and open source software, licensed under GPL-3.

msaenet's Issues

Rewrite parallelization logic

As in 2024, it would be ideal to switch to the %dofuture% operator so that we can

• Reduce boilerplate code.
• Use parallel-safe RNGs.
• Choose from a robust and flexible selection of parallel backends.

https://github.com/nanxstats/r-future-recipes

coef.msaenet() exclues intercept term

I don't know if this is intended or not, but coef.msaenet() excludes the intercept term while coef.glmnet() does not.

library(msaenet)

dat = msaenet.sim.gaussian(
  n = 150, p = 20, rho = 0.6,
  coef = rep(1, 5), snr = 2, p.train = 0.7,
  seed = 1001)

msaenet.fit = msaenet(
  dat$x.tr, dat$y.tr,
  alphas = seq(0.2, 0.8, 0.2),
  nsteps = 3L, seed = 1003)

coef(msaenet.fit)     # coef.msaenet() exclues intercept
coef(msaenet.fit$model)     # coef.glmnet() returns intercept

Release msaenet 3.1.3

Prepare for release

• Check current CRAN check results
• Check licensing of included files
• Review pkgdown reference index for, e.g., missing topics
• Bump version
• Update cran-comments.md (optional)
• Update NEWS.md
• Review pkgdown website
• urlchecker::url_check()
• Check with local machine
• Check with GitHub Actions
• Check with win-builder

Submit to CRAN

• Draft GitHub release
• Submit to CRAN via web form
• Approve emails

Wait for CRAN

• Accepted 🎉

Possibility to provide user-supplied lambda sequence like glmnet

glmnet provides a lambda sequence option for the user to provide their own rather than allowing the algorithm to develop one. Is this a possible feature that could be added in? Or is it a limitation with the adaptive elastic-net?

Thanks.

Release msaenet 3.1.1

Prepare for release

• Check current CRAN check results
• Check licensing of included files
• Review pkgdown reference index for, e.g., missing topics
• Bump version
• Update cran-comments.md (optional)
• Update NEWS.md
• Review pkgdown website
• urlchecker::url_check()
• Check with local machine
• Check with GitHub Actions
• Check with win-builder

Submit to CRAN

• Draft GitHub release
• Submit to CRAN via web form
• Approve emails

Wait for CRAN

• Accepted 🎉

Update the default color palette used in plots

• Update the path plot line colors with the new Tableau 10 color palette.
• Update dot colors in the dotplot.

Can I compute p value using the library msaenet?

Can I compute p-value using the library msaenet?

Support penalty.factor in the initial estimation step for must-keep variables

Under the context of clinical research, it sometimes makes sense to have some "must-keep" variables for Cox regressions. This is beyond the standard penalized linear models supported in msaenet and obviously not officially supported by glmnet or ncvreg.

As a workaround, we can add a new argument penalty.factor.init which will be assigned as the penalty.factor in the first estimation step (and first estimation step only). Users can assign a lower penalty factor for must-keep variables and a higher penalty factor for the other variables. The factor sizes will be subjective, though.

question(coordinate descent)

Hello,did the coordinate descent method be used to estimate the coefficients of elastic network and adaptive elastic network?

Fix alt text for vignette and readme images

As is suggested by the accessibility hints in pkgdown 2.1.0.

Parameter tuning guide vignette

It would be nice to provide a basic guideline vignette for parameter tuning.

Add citation key

Need to add citation key to inst/CITATION so that running citation("msaenet") gives the correct output.

Feature Request - Constraining Covariates

It would be useful if the functions in this package supported or provided arguments similar to glmnet::glmnet()'s lower.limits and upper.limits. I frequently run adaptive elastic nets using glmnet and regularly have to constrain covariates to be non-negative or non-positive.

Release msaenet 3.1.2

Prepare for release

• Check current CRAN check results
• Check licensing of included files
• Review pkgdown reference index for, e.g., missing topics
• Bump version
• Update cran-comments.md (optional)
• Update NEWS.md
• Review pkgdown website
• urlchecker::url_check()
• Check with local machine
• Check with GitHub Actions
• Check with win-builder

Submit to CRAN

• Draft GitHub release
• Submit to CRAN via web form
• Approve emails

Wait for CRAN

• Accepted 🎉

Fix ncvreg warnings on pathwise optimization

Warning: ncvreg() is intended for pathwise optimization, not for single values of lambda.
  1. You are strongly encouraged to fit a path and extract the solution at the lambda value of interest,
      rather than use ncvreg() in this way.
  3. In particular, if you are using the MCP or SCAD penalties, be aware that you greatly increase
      your risk of converging to an inferior local maximum if you do not fit an entire path.
  5. You may wish to look at the ncvfit() function, which is intended for non-path (i.e., single-lambda)
      optimization and allows the user to supply initial values.

no msaenet.sim.gaussian funtion

Use Observable 10 as default palette

https://nanx.me/blog/post/ggsci-3-2-0/