Quantile lasso on a tau by lambda grid — quantile_lasso

Convenience function for computing quantile lasso solutions on a tau by lambda grid.

quantile_lasso_grid(
  x,
  y,
  tau,
  lambda = NULL,
  nlambda = 30,
  lambda_min_ratio = 0.001,
  weights = NULL,
  no_pen_vars = c(),
  intercept = TRUE,
  standardize = TRUE,
  lb = -Inf,
  ub = Inf,
  lp_solver = c("glpk", "gurobi"),
  time_limit = NULL,
  warm_starts = TRUE,
  params = list(),
  transform = NULL,
  inv_trans = NULL,
  jitter = NULL,
  verbose = FALSE
)

Arguments

nlambda: Number of lambda values to consider, for each quantile level. Default is 30.
lambda_min_ratio: Ratio of the minimum to maximum lambda value, for each quantile levels. Default is 1e-3.

Details

This function forms a lambda vector either determined by the nlambda and lambda_min_ratio arguments, or the lambda argument; if the latter is specified, then it takes priority. Then, for each i and j, we solve a separate quantile lasso problem at quantile level tau[i] and tuning parameter value lambda[j], using the quantile_lasso function. All arguments (aside from nlambda and lambda_min_ratio) are as in the latter function; noncrossing constraints are disallowed. The associated predict function is just that for the quantile_genlasso_grid class.