This package presents a method combining Hierarchical Clustering and Group-lasso. Usually, a single partition of the covariates is used in the group-lasso. Here, we provide several partitions from the hierarchical tree.
A post-treatment method based on statistical test (with FWER and FDR control) for selecting the regularization parameter and the optimal group for this value is provided. This method can be applied for the classical group-lasso and our method.
The MLGL function performs the hierarchical clustering and the group-lasso. The post-treatment method can be performed with hierarchicalFWER and selFWER functions. The whole process can be run with the fullProcess function.
"MLGL: An R package implementing correlated variable selection by hierarchical clustering and group-Lasso.", Quentin Grimonprez, Samuel Blanck, Alain Celisse, Guillemette Marot (2018). https://hal.inria.fr/hal-01857242
MLGL, cv.MLGL, fullProcess, hierarchicalFWER
# Simulate gaussian data with block-diagonal variance matrix containing 12 blocks of size 5 X <- simuBlockGaussian(50, 12, 5, 0.7) # Generate a response variable y <- X[, c(2, 7, 12)] %*% c(2, 2, -2) + rnorm(50, 0, 0.5) # Apply MLGL method res <- MLGL(X, y)
Add the following code to your website.
For more information on customizing the embed code, read Embedding Snippets.