R help - Sep 2008 - 1-SE rule in mvpart

Hello,

I'm using mvpart option xv="1se" to compute a regression tree of
good size
with the 1-SE rule.
To better understand 1-SE rule, I took a look on its coding in mvpart, which
is :

Let z be a rpart object ,
xerror <- z$cptable[, 4]
xstd <- z$cptable[, 5]
splt <- min(seq(along = xerror)[xerror <= min(xerror) + xvse * xstd])

I interprete this as following: the simplest tree with xerror under
min(xerror) + its own xstd

Neverthless, in some article I read the following rule:

the simplest tree with xerror under min(xerror) + xstd corresponding to the
min(xerror)

Is this a mistake or any variant of 1-SE rule ? 

-- 
View this message in context:
http://www.nabble.com/1-SE-rule-in-mvpart-tp19508242p19508242.html
Sent from the R help mailing list archive at Nabble.com.

R-users
E-mail: r-help@r-project.org

 Hi!
>I interprete this as following: the simplest tree with xerror under
>min(xerror) + its own xstd
>Neverthless, in some article I read the following rule:
>the simplest tree with xerror under min(xerror) + xstd corresponding to the
>min(xerror)
>Is this a mistake or any variant of 1-SE rule ?
Please refer to page 80 of
Classification and Regression Trees
by Leo Breiman, Jerome Friedman, Charles J. Stone, and R.A. Olshen.
It says

the simplest tree with xerror under min(xerror) + xstd corresponding to the
min(xerror).

although the difference of the two definition is not so large in most cases.
-- 
*****    r.otasuke@gmail.com    *****
http://cse.naro.affrc.go.jp/takezawa/intro.html

	[[alternative HTML version deleted]]