[ADMB Users] estimation of random effects correlations

[Kristian Kleinke]

Hi,

just a short question I've been trying to find an answer to:
am I correct, that glmmadmb() cannot estimate correlations between random effects?

Thanks,
Kristian
_______________________________________________
Users mailing list
Us...@admb-project.org
http://lists.admb-project.org/mailman/listinfo/users

[Mollie Brooks]

Hi Kristian,

The default is to fit uncorrelated random effects (a diagonal variance-covariance matrix), but it appears that you can fit them with a full variance-covariance matrix using the corStruct argument. I can not testify to how well this has been tested.

corStruct

a string specifying the covariance structure of the random effects vector. Two types of covariance matrices are currently implemented: "diag" (diagonal matrix) and "full" (positive definite matrix with all elements being estimated).

cheers,

Mollie

Mollie Brooks, PhD
Postdoctoral Researcher, Population Ecology Research Group, Ozgul Lab, http://www.popecol.org
Institute of Evolutionary Biology & Environmental Studies, University of Zürich

[Ben Bolker]

On 13-05-07 03:46 AM, Mollie Brooks wrote:
> Hi Kristian,
>
> The default is to fit uncorrelated random effects (a diagonal
> variance-covariance matrix), but it appears that you can fit them with a
> full variance-covariance matrix using the corStruct argument. I can not
> testify to how well this has been tested.
>
>
> |corStruct|
>
> a string specifying the covariance structure of the random effects
> vector. Two types of covariance matrices are currently
> implemented: |"diag"| (diagonal matrix) and |"full"| (positive definite
> matrix with all elements being estimated).

This is **NOT** working yet. I'm hoping to have it working soon.

Ben

>
>
> cheers,
> Mollie
>
> Mollie Brooks, PhD
> Postdoctoral Researcher, Population Ecology Research Group, Ozgul
> Lab, http://www.popecol.org
> Institute of Evolutionary Biology & Environmental Studies, University of
> Zürich
>
>
>
> On 7 May 2013, at 9:39 AM, Kristian Kleinke <kmkl...@gmail.com

> <mailto:kmkl...@gmail.com>> wrote:
>
>> Hi,
>>
>> just a short question I've been trying to find an answer to:
>> am I correct, that glmmadmb() cannot estimate correlations between
>> random effects?
>>
>> Thanks,
>> Kristian
>> _______________________________________________
>> Users mailing list

>> Us...@admb-project.org <mailto:Us...@admb-project.org>
>> http://lists.admb-project.org/mailman/listinfo/users