l1svr: Support vector regression and inference under l1...
In jkcshea/l1svr: Inference for Support Vector Regression Under l1 Regularization

Description Usage Arguments Value

View source: R/svr-functions.R

This function performs the support vector regression under l1 regularization. A regression rankscore test is performed to conduct inference and construct confidence intervals.

l1svr(formula, data, epsilon, lambda, h = NULL, kappa = NULL,
  solver = "gurobi", inference = TRUE, confidence.level = NULL,
  confidence.iter = 25, confidence.tol = 0.001,
  confidence.same = 1e-06)

`formula`	Regression formula, e.g., `y ~ x1 + x2`.
`data`	Data frame.
`epsilon`	Scalar defining soft-thresholding rule.
`lambda`	Scalar parameter to scale l1 penalty.
`h`	Scalar determining the quantiles of a standard normal distribution used to define the bandwidth for nonparametric density estimation. If no argument is provided, then errors are assumed to be homoskedastic, and error density estimation is not performed. See Powell (1991) and Bai et. al. (2021) for details.
`kappa`	Scalar that directly scales the bandwidth for nonparametric density estimation. If no argument is provided, then errors are assumed to be homoskedastic, and error density estimation is not performed. See Powell (1991) and Bai et. al. (2021) for details.
`solver`	Character, name of the linear programming package in R used to obtain the bounds on the treatment effect. The function supports `'gurobi'`, `'cplexapi'`, `'lpsolveapi'`. The name of the solver should be provided with quotation marks.
`inference`	Boolean, determines whether or not p-values and confidence intervals will be estimated. Set to `TRUE` by default.
`confidence.level`	Scalar between 0 and 1. Determines the confidence level for estimating confidence intervals.
`confidence.iter`	Integer determining maximum number of iterations performed when estimating the confidence interval. Set to 25 by default.
`confidence.tol`	Scalar between 0 and 1. Determines the tolerance for the confidence level when estimating the confidence interval.
`confidence.same`	Scalar between 0 and 1. Sets the tolerance for determining when the confidence interval estimate may no longer be improved.