fit.adaptive.hlasso: Fit adaptive hierarchical lasso

In denisagniel/smrtr:

Fit adaptive hierarchical lasso

Description

Fit adaptive hierarchical lasso

Usage

fit.adaptive.hlasso(x.t, y, beta.ini, m, n, p, info.half, BIC = FALSE,
  max.iter = 50, print = FALSE)

Arguments

x.t	Design matrix or pseudo-design matrix where each column has been multiplied by abs(beta) corresponding to it
y	Outcome or pseudo-outcome matrix
m	Number of outcomes
n	Number of observations
p	Number of predictors
info.half	Half Fisher information matrix for beta
BIC	Logical, if FALSE, modified BIC criterion is used
max.iter	Maximum number of iterations in iterative procedure
print	Logical indicating whether to print intermediate updates

Details

Fits an adaptive version of the hierarchical lasso suitable for multivariate mixed outcomes. Variable selection is performed to both remove uninformative predictors and remove unimportant predictor-outcome relationships. Tuning parameter selection is done via a BIC-style criterion.

Value

A list with the following elements:

• lasso.beta: the non-hierarchical adaptive lasso estimate

• hlasso.beta: the adaptive hierarchical lasso estimate

• num.iter: the number of iterations used

• d: the estimate of d

• alpha: the estimate of alpha

See Also

lars

denisagniel/smrtr documentation built on Sept. 17, 2022, 10:23 p.m.