[S] Problem using edit.tree() in tree-based regression

Gunter, Bert (bert_gunter@merck.com)
Thu, 19 Feb 1998 08:16:12 -0500


This question is for those interested in/familiar with S implementation
of regression trees via tree(). (NT4.0 platform, S-Plus 4, latest
update).

I use burl.tree() to look for alternative splits, often on trees
obtained from prune.tree() and snip.tree(), not the original (overfit)
tree. I then want to use edit.tree() to see how these alternatives do.
However, on the objects returned by anything but tree(), I get the
error:

Error in call to edit.tree(): Argument subset= not matched
Dumped

Can someone tell me what's happening here and provide me a correction or
workaround. The obvious alternative of giving the output of tree() as
the argument to edit.tree is fine if you only wish to change one node,
but what if you want to change more than one? That is, using burl.tree()
on an edit.tree() output (or prune.tree(edit.tree()) output) I might
find more alternative splits to examine, but I can't seem to use
edit.tree on this object.

Please reply to me personally and I shall post a summary.

-----------------------------------------------------------------------
This message was distributed by s-news@wubios.wustl.edu. To unsubscribe
send e-mail to s-news-request@wubios.wustl.edu with the BODY of the
message: unsubscribe s-news