R/helper_functions.R

In hdme: High-Dimensional Regression with Measurement Error

.onUnload <- function (libpath) {
  library.dynam.unload("hdme", libpath)
}

# Logistic functions
logit <- stats::plogis
dlogit <- function(x) exp(-x)*(1+exp(-x))^(-2)

# Poisson functions
pois <- function(x) exp(x)
dpois <- function(x) exp(x)



set_radius <- function(W, y, family = "gaussian", no_radii,
                       limit_factors = c(2e-3, 2)) {

  no_radii <- ifelse(is.null(no_radii), 20, no_radii)

  # First run the naive Lasso
  lassoFit <- glmnet::cv.glmnet(W, y, family = family)
  betaNaive <- stats::coef(lassoFit, s = "lambda.min")

  # Use the estimated vector to find the upper radius for cross-validation
  a <- sum( abs( betaNaive ) )
  # Set the cross-validation range
  radius <- seq(from = limit_factors[1] * a, to = limit_factors[2] * a,
                length.out = no_radii)

}


set_up_cv <- function(N, n_folds){
  list(
    fold_id = sample(rep(seq(n_folds), length = N)),
    outlist = as.list(seq(n_folds))
  )
}