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: xclust
See also: dentoreg cartsplit cartsplitopt cartcv leafnum maketr pred prederr prune prunecv pruneseq prunetot ssr kuva

Macro: denvalues
Description: Given a binary tree produced by cartsplit, normalises the mean values of the leaves so that the function represented by the binary tree integrates to one.

Usage: val = denvalues (tr, dendat)
Input:
tr list of vectors: data structure which represents a binary tree and is produced by cartsplit procedure, contains vectors tr.val, tr.vec, tr.mean, tr.ssr, tr.nelem. See cartsplit for the description of tr.
dendat n x p matrix: data matrix of i.i.d. observations. Data with which the density tree was estimated.
Output:
val vector which has as many elements as there are elements in the vectors which form the tree tr. In the positions which correspond to the leaf nodes there are normalised values of vector tr.mean. Values are normalised in such a way that the function represented by the binary tree integrates to one. In other, nonterminal positions, there are Inf.

Note:

Example:
; loads the library xclust
library ("xclust")
; generate the density data
randomize(1)
dendat=uniform(10)~(uniform(10)+1)
; transform the density data to a regression data
regdat=dentoreg(dendat,4)
; estimate the density tree
cs=cartsplit(regdat.ind,regdat.dep,#(1,1))
; normalise the density estimate
cval=denvalues(cs,dendat)
cval
cs.mean=cval
Result:

Contents of cval
[ 1,] Inf 
[ 2,] Inf 
[ 3,] Inf 
[ 4,] 0.183560 
[ 5,] 0.351490 
[ 6,] Inf 
[ 7,] 0.461948 
[ 8,] 0.550679 
[ 9,] Inf 
[10,] Inf 
[11,] Inf 
[12,] 0.183560 
[13,] 0.183560 
[14,] Inf 
[15,] 0.183560 
[16,] 0.351490 
[17,] Inf 
[18,] 0.461948 
[19,] Inf 
[20,] 0.183560 
[21,] 0.351490 

Library: xclust
See also: dentoreg cartsplit cartsplitopt cartcv leafnum maketr pred prederr prune prunecv pruneseq prunetot ssr kuva

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

Author: Jussi Klemelae, 980323
(C) MD*TECH Method and Data Technologies, 28.6.1999