KIR: Kernel Inverse Regression
In abnormally-distributed/cvreg: Cross Validation and Robust Estimation Utilities

Description Usage Arguments Value References

Kernel inverse regression (KIR) is very similar to PIR (Zhu & Fang, 1996). Instead of using a polynomial representation of Y, Y is represented through a kernel function, k. The inverse regression E[X|Y] is calculted as E[Xk(Y-y)] / E[k(Y-y)]. The covariance of E[X|Y] is calculated as the average values of the transposed crossproduct of E[X|Y]-E[X], ie, Sigma = E[E[X|Y]-E[X] * E[X|Y]-E[X]']. The effective dimension reduction subspace is then estimated through the generalized eigenvalue problem GEV(Sigma, cov(X)).

KIR(
  formula,
  data,
  rank = "all",
  kern = c("rbf", "gt", "cauchy", "spline", "laplace", "bessel", "anova"),
  sigma = "auto",
  order = 1,
  df = 3,
  lambda = 1e-04
)

`formula`	a model formula
`data`	a data frame
`rank`	the desired number of sufficient predictors to return. the default is "all".
`kern`	the kernel function to be used. must be one of "rbf", (the default), "gt, "cauchy", "spline", "laplace", "bessel", or "anova".
`sigma`	the scale to be used for the kernel. defaults to "auto" which estimates the optimal sigma value using kernlab::sigest.
`order`	the order to be used for the bessel function.
`df`	the degrees of freedom to be used for the bessel kernel, anova kernel, and generalized t kernel. defaults to 3.
`lambda`	a tuning parameter for regularization. defaults to 1e-4.