XploRe Help : sir1

Keywords - Function groups - @ A B C D E F G H I J K L M N O P Q R S T U V W X Y Z

Library:	metrics
See also:	sssm sir sir2

Quantlet:		sir1
Description:		the taylored Sliced Inverse Regression method for the "sssm" quantlet.

Usage:	`res = sir1(matYX,nbslices)`
Input:
	`matYX`	n x (p+1) matrix, first column contains values of the dependent variable y, next p columns contain the values of the p explanatory variables; n stands for the sample size.
	`nbslices`	optional scalar, number of slices for the non-missing y's cases. Default: nbslices = 5.
Output:
	`res.edr`	p x p matrix, the estimated EDR directions (columnwise), sorted according to the increasing values of the corresponding eigenvalues.
	`res.eigen`	p x 1 vector, the corresponding eigenvalues sorted according to the increasing values.

Note:

"sir1" builds a special slice which contains the cases such that y==0 (the value "0" for y symbolically indicates a missing value). Then, nbslices (by default, nbslices=5) other slices are made by splitting the range of the non-missing y's into slices of nearly equal weight (i.e. such that they contain nearly the same number of cases). The classical algorithm of S.I.R. is then used with this special slicing.

Author: J. Saracco, 20030127 license MD*tech