[R] how to put n and yprob on the tree plot ----- Re: what missed ----- CART

Uwe Ligges ligges at statistik.tu-dortmund.de
Sun Feb 24 22:44:37 CET 2008



o ha wang wrote:
> Thanks!
>    
>   Yes, you were right. I loaded package 'tree'.
>    
>   I tried 'tree' and 'rpart' respectively, it looks like results for regression tree are same, and very similar for classification tree. (more biref using 'rpart' than using 'tree').
>    
>   But my problems are still not been solved using rpart.
>    
>   1. how to put number (n, ie. class counts) and probability (yprob) on the plot for classification tree. There are only 0/1 on the plot.  n and yprob can be seen using 'myfit' but not on the plot.
>    
>   2. The notation of Numbers are overlapped for regression tree. 
>    
>   please see my rpart code:
>    
>   library (rpart) 
> myfit <- rpart (y ~ x1 + x2 + x3 + x4 ) 
> plot (myfit)); text (myfit, all= TRUE, cex=0.5, pretty=0)
> myfit
>   summary(myfit)


I still cannot reproduce without y, x1, x2, x3, x4

Uwe Ligges


> thanks, xiao yue
>    
>   
> Uwe Ligges <ligges at statistik.tu-dortmund.de> wrote:
>   
> 
> o ha wang wrote:
>> Hi all,
>>
>> Can anyone who is familar with CART tell me what I missed in my tree code? 
>>
>> library (MASS) 
>> myfit <- tree (y ~ x1 + x2 + x3 + x4 ) 
> 
> There is not function tree() in MASS. I guess you have loaded package 
> "tree"?
> Note that the author and maintainer of the tree package suggests to use 
> package rpart instead.
> 
> 
> 
>> # tree.screens () # useless
>> plot(myfit); text (myfit, all= TRUE, cex=0.5, pretty=0)
>> # tile.tree (myfit, fgl$type) # useless
>> # close.screen (all= TRUE) # useless
>>
>>
>> My current tree plot resulted from above code shows as: 
>> 1. overlapped #s caused by unsuitable length of branch.
>> 2. no misclassification rates: 'misclass.tree' only brings up the error of ' misclassification error rate is appropriate for factor responses only', but my response y is 0/1 data.
> 
> If you data is 0/1, you should make it a factor otherwise the cose will 
> assume you want a regression rather than a classification tree.
> 
> 
>> 3. Unsuitable location of notations: there are not two notation of splitting criteria on the two branches when a node is split, instead only one notation of splitting criteria is on the node location. 
> 
> It indicates which items go into the left branch.
> 
> Uwe Ligges
> 
> 
>> thanks,
>> xiao yue
>>
>>
>> ---------------------------------
>>
>> [[alternative HTML version deleted]]
>>
>> ______________________________________________
>> 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.
> 
> 
>        
> ---------------------------------
> 
> 	[[alternative HTML version deleted]]
> 
> ______________________________________________
> 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.



More information about the R-help mailing list