[R] mob(party) formula question

[Achim Zeileis]

On Wed, 13 Aug 2008, Birgitle wrote:

> Thanks again.
> Unfortunately I have always this missing values problem.
> But the missings have also a meaning and its impossible to code it
> differently or impute.

That's ok. Just to clarify: NAs are not allowed in the response or the 
modeling variables. In principle, it would be possible to have NAs in the 
partitioning variables and try to handle it with surrogate splits. 
Currently, surrogates are not implemented in mob(), but we are currently 
working on infrastructure for this.

So the only work-around easily available at the moment is to call 
na.omit() (on the relevant variables only).

Best,
Z

> Also thanks for the explanation. Now I understand.
>
> B.
>
>
> Achim Zeileis wrote:
>>
>> On Wed, 13 Aug 2008, Birgitle wrote:
>>
>>> Many thanks for your answer and the code that you offered me.
>>>
>>> I get this error message after calling mob (look at my given example).
>>> I guess it has something to do with the missings?
>>
>> Yes, you have to handle NAs in advance if you want to fit that model.
>> We'll try to fix that in future versions.
>>
>>> The iris example works also fine for me.
>>>
>>> Sorry that I am not enough into statistics to really understand the
>>> following:
>>>
>>>
>>> Achim Zeileis wrote:
>>>>
>>>>
>>>> .....
>>>> For the variables for which a linear specification makes sense (at least
>>>> in each component) then you should include them for modeling. And those
>>>> variables for which it is not clear a priori what a useful parametric
>>>> specification would be should be used as partitioning variables.
>>>> ...
>>>>
>>>>
>>>
>>> What do you mean with "linear specification"? I would be very happy if
>>> you
>>> could explain.
>>
>> Well, in each node you fit a logistic regression model. This is a
>> (generalized) linear model, hence the variables included have a linear
>> influence (on the link scale) within each node. The partitioning variables
>> on the other hand capture step-shaped influences (if they are selected by
>> the algorithm). See the references on ?mob for further details.
>>
>> Best,
>> Z
>>
>> ______________________________________________
>> 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.
>>
>>
>
>
> -----
> The art of living is more like wrestling than dancing.
> (Marcus Aurelius)
> -- 
> View this message in context: http://www.nabble.com/mob%28party%29-formula-question-tp18959898p18964864.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.
>
>