search

topepo / FES

Code and Resources for "Feature Engineering and Selection: A Practical Approach for Predictive Models" by Kuhn and Johnson
https://bookdown.org/max/FES
GNU General Public License v2.0
724 stars 237 forks source link

Some ideas for interactions chapter #40

Closed LluisRamon closed 5 years ago

LluisRamon commented 6 years ago

In Chapter 7 Detecting Interaction Effects and particularly in Section 7.4.2 The Lasso, I think that it could be interesting to provide some comments on

Michael Lim & Trevor Hastie 2016 Learning interactions via hierarchical group-lasso regularization

The authors provide a package with their method in CRAN called glinternet.

It was compared in

A systematic comparison of statistical methods to detect interactions in exposome-health associations. https://www.ncbi.nlm.nih.gov/pubmed/28709428

and glinternet was one of the methods with better performance.

pejovic commented 6 years ago

Dear Max and Kjell

In relation to your book, we have recently published the paper "Sparse regression interaction models for spatial prediction of soil properties in 3D", which you could find interesting and useful. It can be related to the Chapter 7 "Detecting interaction effects", especially with the sub-Chapter 7.4.2. "lasso". In our paper, we demonstrated use of lasso and hierarchical lasso for detecting important interactions for spatial prediction of soil variables at different soil depths. One of the conclusions was that 3D prediction of soil properties using linear models benefits from interaction effects among spatial predictors and soil depth only if detection of important interactions is done via lasso. Modeling and evaluating procedures are given in the paper and the approach is demonstrated on three case studies and compared to other linear modeling approaches (OLS and stepwise regression). We'll be glad if you find our paper useful.

Link to the paper: https://www.sciencedirect.com/science/article/pii/S009830041730852X

topepo commented 5 years ago

@LluisRamon

In Chapter 7 Detecting Interaction Effects and particularly in Section 7.4.2 The Lasso, I think that it could be interesting to provide some comments on

Michael Lim & Trevor Hastie 2016 Learning interactions via hierarchical group-lasso regularization

The authors provide a package with their method in CRAN called glinternet.

It is a really interesting paper that I was eager to test out. Unfortunately, the user interface to that R package is so bad, I feel that it would be irresponsible to use it. Those are strong words, but I don't think that the average R user would be able to successfully navigate their data input requirements or the format of their output. It is very frustrating.

LluisRamon commented 5 years ago

Thanks for considering, it was a suggestion and your concerns are pretty reasonable.