[R] R - rpart - increasing xerror

[Uwe Ligges]

On 18.08.2012 21:32, Daniel Blankenheim wrote:
> Hey
> my name is daniel,  i am writing my bachelor thesis and wondering if you can
> help me.
> i am trying to generate a regression tree via  rpart. to
> reduce the error of the model i use cross validation, but instead
> of reducing the cross validation error (xerror) is increasing the more splits
> there are.

... which indicates overfitting. If y is independent of you x(s) the 
best prediction is the overall mean, and that happens with 0 splits.
Anyway, this is somthing to ask your supervisor, since this list is 
about R-help rather than help on statistical modelling.

Best,
Uwe Ligges








> I dont understand what that means. does it mean that my model doesnt fit the
> data- that there is no trend in the data? i am 100% sure that
> there must be a trend or correlation in the data.
>
>>> please help me:)
>
>>>          CP nsplit rel error xerror    xstd
>>> 1 0.100022      0   1.00000 1.0192 0.14222
>>> 2 0.066716      2   0.79996 1.3107 0.18720
>>> 3 0.050471      3   0.73324 1.4127 0.21138
>>> 4 0.033758      4   0.68277 1.5197 0.22826
>>> 5 0.010376      5   0.64901 1.5360 0.23792
>>> 6 0.000010      6   0.63864 1.5419 0.24280
>
> ______________________________________________
> 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.
>