Frequently Asked Questions¶
Here we’ll maintain a list of frequently asked questions about groupyr. Do you have a question that isn’t addressed here? If so, please see our getting help page for information about how to file a new issue.
Why did we create groupyr and how does it compare to other similar packages?
We created groupyr to be useful in our own research and we hope it is
useful in yours. There are other packages for penalized regression in
python. The lightning
package has a lasso penalty but does not allow the user to specify
groups. The group_lasso package is well
designed and documented, but we found that groupyr’s execution time was
faster for most problems. We also wanted estimators with built-in
cross-validation using both grid search and the
sequential model based optimization.
In the future, we hope that groupyr will include other methods for statistical learning with grouped covariates (e.g. unsupervised learning methods). These would also be out of scope for the aforementioned libraries. However, we encourage you to try many tools. If you find that another one is better suited to your problem, please leave us some feedback, go forth, and do good work. Happy coding!