[R-sig-ME] How to extract num. of observations and groups used in a mixed model using gee package

[Robert Kushler]

The following lines from the gee help page ...

Details

Though input data need not be sorted by the variable named "id", the program will interpret physically contiguous 
records possessing the same value of id as members of the same cluster. Thus it is possible to use the following vector 
as an id vector to discriminate 4 clusters of size 4: c(0,0,0,0,1,1,1,1,0,0,0,0,1,1,1,1).


  ... indicate that the proposed strategy may not yield a correct answer.  In fact, the above suggests
that you could get different results depending on whether or not your data set is sorted.  I find
this quite disturbing, but haven't taken the time to test it.

Regards,   Rob Kushler





On 4/24/2012 10:25 AM, David Winsemius wrote:
>
> On Apr 24, 2012, at 9:46 AM, Ben Bolker wrote:
>
>> david oseguera montiel <oseguera.david at ...> writes:
>>
>>>
>>> I meant the grouping factor in the model indicated by id. From gee
>>> documentation 'id = a vector which identifies the clusters..'. So in the
>>> warpbreaks data set, id = wool has 54 obs, in two clusters (groups).
>>
>> [snip]
>>
>>>>> David,
>>>>> Thank you very much. Number of observations solved, but I am still
>>>>> missing how to find out the number of groups used. In the warpbreaks
>>>>> data set these are 2. Any ideas. Thank you.
>>>>>
>>>> I must be unclear what you mean by "number of groups".
>>
>>
>> Didn't Chuck Cleland's answer work, i.e. length(unique(model$id)) ... ?
>
> It didn't work for me.
> length(unique(summgee$id))
> [1] 0
>  > summgee$id
> NULL
>
> See my other reply for what did work. I suppose it could be that both the OP and I are behind on gee versions. Mine is
> 4.13-17 and I see that there is a more recent binary on CRAN. With a (slightly out-of-date) version loaded, I am loath
> to detach because getting things to all sync up is a bit of a hassle and error-prone without a fresh start and I have
> other work time-consuming running in this session.
>