sir: sir models

In alex-conanec/OptFilBov: Compromise dealing

Description

sir is a method used to fit sliced-inverse regression models

Usage

sir(Y, X, H = 10, K = 1, Ylabel = NULL)

Arguments

Y	numerical vector of size n containing the response to predict.
X	numerical matrix of dimension n by p regressors
H	integer (by default 10) which set the number of slices
K	integer (by default 1) which set the number of component taken. Only 1 is possible right now.
Ylabel	a character value containing the name of the Y varaible

Value

An object with S3 class "sir" with the following components:

beta
index
H	H parameter passed througth the function
K	K parameter passed througth the function
hopt	bandwidth value of the kernel tuned to smooth the points and predict new values. For more explaination see cv_bandwidth
Xlabels	a character vector containing the p regessor names
Ylabel	a character value containing the response variable name
x_train	the training matrix X of dimension p times n
y_train	the training response vector Y of length n

See Also

predict.sir

Examples

library(mfe)
data(indicateurs)
X <- indicateurs[, -c(1,2,3)]
Y <- indicateurs[,1]
model <- sir(X = X, Y = Y, Ylabel = colnames(indicateurs)[1])
predict(model, newdata = indicateurs[1,])

alex-conanec/OptFilBov documentation built on May 21, 2019, 9:46 a.m.