R help - May 2012 - rpart - predict terminal nodes for new observations

Dear useRs:

Is there a way I could predict the terminal node associated with a new data
entry in an rpart environment? In the example below, if I had a new data
entry with an AM of 5, I would like to link it to the terminal node 2. My
searches led to http://tolstoy.newcastle.edu.au/R/e4/help/08/07/17702.html
but I do not seem to be able to operationalize Professor Ripley's
suggestions.

Many thanks.

Tudor
> tree.prune
n= 2400 

node), split, n, loss, yval, (yprob)
      * denotes terminal node

1) root 2400 779 0 (0.6754167 0.3245833)  
  2) AM< 6.5 1428 254 0 (0.8221289 0.1778711) *
  3) AM>=6.5 972 447 1 (0.4598765 0.5401235)  
    6) P>=10.39666 390  86 0 (0.7794872 0.2205128) *
    7) P< 10.39666 582 143 1 (0.2457045 0.7542955) *


--
View this message in context:
http://r.789695.n4.nabble.com/rpart-predict-terminal-nodes-for-new-observations-tp4630104.html
Sent from the R help mailing list archive at Nabble.com.

On 15.05.2012 16:30, tudor wrote:
> Dear useRs:
>
> Is there a way I could predict the terminal node associated with a new data
> entry in an rpart environment? In the example below, if I had a new data
> entry with an AM of 5, I would like to link it to the terminal node 2. My
> searches led to http://tolstoy.newcastle.edu.au/R/e4/help/08/07/17702.html
> but I do not seem to be able to operationalize Professor Ripley's
> suggestions.

Use the predict() function.

Uwe Ligges

> Many thanks.
>
> Tudor
>
>> tree.prune
> n= 2400
>
> node), split, n, loss, yval, (yprob)
>        * denotes terminal node
>
> 1) root 2400 779 0 (0.6754167 0.3245833)
>    2) AM<  6.5 1428 254 0 (0.8221289 0.1778711) *
>    3) AM>=6.5 972 447 1 (0.4598765 0.5401235)
>      6) P>=10.39666 390  86 0 (0.7794872 0.2205128) *
>      7) P<  10.39666 582 143 1 (0.2457045 0.7542955) *
>
>
> --
> View this message in context:
http://r.789695.n4.nabble.com/rpart-predict-terminal-nodes-for-new-observations-tp4630104.html
> Sent from the R help mailing list archive at Nabble.com.
>
> ______________________________________________
> R-help at r-project.org mailing list
> https://stat.ethz.ch/mailman/listinfo/r-help
> PLEASE do read the posting guide
http://www.R-project.org/posting-guide.html
> and provide commented, minimal, self-contained, reproducible code.

Hi Uwe:

Thanks a lot for your reply.  

I contemplated the use of the predict() function but realized that it could
not return what I needed (i.e., the terminal nodes and not the classes). I
went back to the post of Professor Ripley
(http://tolstoy.newcastle.edu.au/R/e4/help/08/07/17702.html) and did find a
way to get the terminal nodes.  I am showing this approach below in case
other useRs face a similar problem.

Many thanks.

Tudor


nodes  <- rpart:::pred.rpart(tree.prune,rpart:::rpart.matrix(es))
where 
tree.prune is the rpart object and es is the estimation data set.

--
View this message in context:
http://r.789695.n4.nabble.com/rpart-predict-terminal-nodes-for-new-observations-tp4630104p4630508.html
Sent from the R help mailing list archive at Nabble.com.