R help - May 2008 - plotcp() in rpart()

I have a general question about how to interpret the plotcp() graph.

The cross-validation "xerror" value typically follows a decreasing
pattern, from approximately 1.0 at the root node, then it crosses the 1SE
boundary, reaches a plateau, and decreases further when the tree gets very
complex [e.g., Venables & Ripley, 4ed, p.260].   The preferred tree is the
one before it crosses the 1SE boundary.

What happens if plotcp() shows an V-shape profile?  For example, it goes down,
reaches the lowest point at a tree of size 6, then comes back up (see attached
pdf graph).  It seems that the predictors (I have a few) do not help the splits.
Would it be reasonable to prune it at size 6?  Or perhaps rpart() is not
suitable for this analysis?

Yuelin.



 
     ====================================================================     
     Please note that this e-mail and any files transmitted with it may be 
     privileged, confidential, and protected from disclosure under 
     applicable law. If the reader of this message is not the intended 
     recipient, or an employee or agent responsible for delivering this 
     message to the intended recipient, you are hereby notified that any 
     reading, dissemination, distribution, copying, or other use of this 
     communication or any of its attachments is strictly prohibited.  If 
     you have received this communication in error, please notify the 
     sender immediately by replying to this message and deleting this 
     message, any attachments, and all copies and backups from your 
     computer.
-------------- next part --------------
A non-text attachment was scrubbed...
Name: Temp1.pdf
Type: application/pdf
Size: 7493 bytes
Desc: not available
URL:
<https://stat.ethz.ch/pipermail/r-help/attachments/20080530/6b6b262d/attachment.pdf>

What happens if plotcp() shows an V-shape profile?  
   It seems that the predictors (I have a few) do not help the splits.  
   Would it be reasonable to prune it at size 6?  
   Or perhaps rpart() is not suitable for this analysis?  

What I meant on the last sentence was perhaps I made a poor choice of the
predictors to enter into rpart.  They do not seem to be associated with the
classes.

Yuelin.
   

 
     ====================================================================     
     Please note that this e-mail and any files transmitted with it may be 
     privileged, confidential, and protected from disclosure under 
     applicable law. If the reader of this message is not the intended 
     recipient, or an employee or agent responsible for delivering this 
     message to the intended recipient, you are hereby notified that any 
     reading, dissemination, distribution, copying, or other use of this 
     communication or any of its attachments is strictly prohibited.  If 
     you have received this communication in error, please notify the 
     sender immediately by replying to this message and deleting this 
     message, any attachments, and all copies and backups from your 
     computer.