dss_select: Group lasso for matrix regression

In drkowal/dfosr: Dynamic Function-on-Scalars Regression

Description

Given a N x M data matrix Y and a N x P matrix of predictor X, estimate the model Y = XB + E for P x M regression coefficient matrix B. The penalty applies a row-wise group lasso penalty, i.e., for row p, the M elements X[p,] are regularized toward zero.

Usage

dss_select(Y, X, w = NULL)

Arguments

Y	N x M matrix of observations
X	N x P matrix of predictors
w	P x 1 vector of weights; if NULL, use sqrt(M)

Value

The solution path for L values of lambda in the form of an array of dimension M x P x L

Note

The design matrix X may include an intercept, which will be left unpenalized.

drkowal/dfosr documentation built on May 7, 2020, 3:09 p.m.