[R-sig-ME] zero-inflated models with spatial autocorrelation

Malcolm Fairbrother M.Fairbrother at bristol.ac.uk
Fri May 24 16:02:17 CEST 2013


Dear Ben and Ricardo,

Can I just clarify: Are you talking about similar functionality to
that in nlme... but, per Ricardo's query, with zero-inflation?

Aside from the issue of zero-inflation, one thing that is limiting
about nlme is that it only allows for spatial autocorrelation at the
lowest level. That is, each observation has to have a unique location.
It would be great to have functionality in lme4 that allows for
observations to be nested in *groupings* that have known locations...
so the locations of two lowest-level observations need not be unique.

Just trying to reinforce Ben's gleam!

Cheers,
Malcolm


> Date: Fri, 24 May 2013 09:31:25 +0100
> From: Ricardo Lima <rfaustinol at gmail.com>
> To: Ben Bolker <bbolker at gmail.com>
> Cc: r-sig-mixed-models at r-project.org
> Subject: Re: [R-sig-ME] zero-inflated models with spatial
>         autocorrelation
>
> Thanks for the quick reply!
>
> Yes, I meant continuous spatial autocorrelation. It is indeed surprising
> that no package in R can do this, when this is such a big issue when it
> comes to analysing ecological data.
>
> Best regards,
> RICARDO.
>
>
>
>
>
> 2013/5/23 Ben Bolker <bbolker at gmail.com>
>
>> Ricardo Lima <rfaustinol at ...> writes:
>>
>> >
>> > Hi,
>> >
>> > Is there any R package that allows to create zero-inflated models with
>> > spatial autocorrelation?
>> >
>> > Best regards,
>> > RICARDO Lima
>>
>>   Not really, if you mean continuous (geostatistical-style) spatial
>> autocorrelation. In principle, you can code it yourself in AD Model
>> Builder, or WinBUGS, or Stan ...
>>
>> http://www.admb-project.org/examples/spatial-models
>>
>>   This (and even the simpler case of a 'proper' GLMM with spatial
>> autocorrelation in a latent observation-level random variable) is
>> a surprisingly missing component in the available R toolbox, as far
>> as I know.  INLA _might_ do the simpler (non-ZI) case.  I have thought
>> about some ways to implement this in the development version of lme4,
>> but so far it's a gleam in my eye.
>>
>>   I would also check out
>>
>> http://www.highstat.com/book4.htm



More information about the R-sig-mixed-models mailing list