pkg/R/computeGridLambda.R

   1 #' computeGridLambda
   2 #'
   3 #' Construct the data-driven grid for the regularization parameters used for the Lasso estimator
   4 #'
   5 #' @param phiInit value for phi
   6 #' @param rhoInit  for rho
   7 #' @param piInit  for pi
   8 #' @param gamInit value for gamma
   9 #' @param X matrix of covariates (of size n*p)
  10 #' @param Y matrix of responses (of size n*m)
  11 #' @param gamma power of weights in the penalty
  12 #' @param mini minimum number of iterations in EM algorithm
  13 #' @param maxi maximum number of iterations in EM algorithm
  14 #' @param eps threshold to stop EM algorithm
  15 #' @param fast boolean to enable or not the C function call
  16 #'
  17 #' @return the grid of regularization parameters for the Lasso estimator. The output is a vector with nonnegative values that are relevant
  18 #' to be considered as regularization parameter as they are equivalent to a 0 in the regression parameter.
  19 #'
  20 #' @export
  21 computeGridLambda <- function(phiInit, rhoInit, piInit, gamInit, X, Y, gamma, mini,
  22   maxi, eps, fast)
  23 {
  24   n <- nrow(X)
  25   p <- ncol(X)
  26   m <- ncol(Y)
  27   k <- length(piInit)
  28
  29   list_EMG <- EMGLLF(phiInit, rhoInit, piInit, gamInit, mini, maxi, gamma, lambda = 0,
  30     X, Y, eps, fast)
  31
  32   grid <- array(0, dim = c(p, m, k))
  33   for (j in 1:p)
  34   {
  35     for (mm in 1:m)
  36       grid[j, mm, ] <- abs(list_EMG$S[j, mm, ])/(n * list_EMG$pi^gamma)
  37   }
  38   sort(unique(grid))
  39 }