Multi-species models

[jaya]

Dear forum,

I just wanted to check to see if any multi-species models have been
implemented in Unmarked yet, or if anyone had any code they could
share to help me with this.

Beggars can't be choosy, but the Yamaura et al. (JAPPL 2011) multi-
species RN model would be especially useful...

Many thanks!

[Richard Chandler]

Hi Jedediah,

These models won't be in unmarked for quite some time I'm afraid. But Dan Linden at MSU is working on it. And I might piece together a function for the workshop in April, but it won't be easy to add to unmarked. In the meantime, you can fit these models using BUGS or JAGS.

Richard
_____________________________________
Richard Chandler, post-doc
USGS Patuxent Wildlife Research Center
301-497-5696

From:	jaya <jededia...@gmail.com>
To:	unmarked <unma...@googlegroups.com>
Date:	03/08/2012 09:28 PM
Subject:	[unmarked] Multi-species models
Sent by:	unma...@googlegroups.com

---

[Manuel Spínola]

Hi Jedediah and Richard,

May be I am misunderstanding the question, but is it possible to use the species as a covariate (factor) in one of the unmarked models?

Best,

Manuel

2012/3/9 Richard Chandler <rcha...@usgs.gov>

--
Manuel Spínola, Ph.D.
Instituto Internacional en Conservación y Manejo de Vida Silvestre
Universidad Nacional
Apartado 1350-3000
Heredia
COSTA RICA
mspi...@una.ac.cr
mspin...@gmail.com
Teléfono: (506) 2277-3598
Fax: (506) 2237-7036
Personal website: Lobito de río
Institutional website: ICOMVIS

[Richard Schuster]

Hi Manuel,

As far as I know that's what some people are doing that use program PRESENCE (e.g. De Wan et al 2009 BiolCons).

Cheers,
Richard S

Hi Jedediah and Richard,

Manuel Spï¿œnola, Ph.D.
Instituto Internacional en Conservaciï¿œn y Manejo de Vida Silvestre

Universidad Nacional
Apartado 1350-3000
Heredia
COSTA RICA
mspi...@una.ac.cr
mspin...@gmail.com

Telï¿œfono: (506) 2277-3598
Fax: (506) 2237-7036
Personal website: Lobito de rï¿œo
Institutional website: ICOMVIS

[Manuel Spínola]

I did it with occupancy models in unmarked but may be it was not appropiate, I don´t know. 

Best,

Manuel

2012/3/9 Richard Schuster <ric.sc...@gmail.com>

Hi Manuel,

As far as I know that's what some people are doing that use program PRESENCE (e.g. De Wan et al 2009 BiolCons).

Cheers,
Richard S

Hi Jedediah and Richard,

Manuel Spínola, Ph.D.
Instituto Internacional en Conservación y Manejo de Vida Silvestre

Universidad Nacional
Apartado 1350-3000
Heredia
COSTA RICA
mspi...@una.ac.cr
mspin...@gmail.com

Teléfono: (506) 2277-3598
Fax: (506) 2237-7036
Personal website: Lobito de río
Institutional website: ICOMVIS

--
Manuel Spínola, Ph.D.
Instituto Internacional en Conservación y Manejo de Vida Silvestre

Universidad Nacional
Apartado 1350-3000
Heredia
COSTA RICA
mspi...@una.ac.cr
mspin...@gmail.com

Teléfono: (506) 2277-3598
Fax: (506) 2237-7036
Personal website: Lobito de río
Institutional website: ICOMVIS

[Richard Schuster]

De Wan also did it using occupancy models and in one of the PRESENCE tutorials and in their phidot.org forum I read about other people doing that too.

Cheers,
Richard S

Am 3/9/2012 07:35, schrieb Manuel Spï¿œnola:

I did it with occupancy models in unmarked but may be it was not appropiate, I don´t know.ï¿œ

Best,

Manuel

2012/3/9 Richard Schuster <ric.sc...@gmail.com>

Hi Manuel,

As far as I know that's what some people are doing that use program PRESENCE (e.g. De Wan et al 2009 BiolCons).

Cheers,
Richard S

Hi Jedediah and Richard,

Manuel Spï¿œnola, Ph.D.
Instituto Internacional en Conservaciï¿œn y Manejo de Vida Silvestre

Universidad Nacional
Apartado 1350-3000
Heredia
COSTA RICA
mspi...@una.ac.cr
mspin...@gmail.com

Telï¿œfono: (506) 2277-3598
Fax: (506) 2237-7036
Personal website: Lobito de rï¿œo
Institutional website: ICOMVIS

--
Manuel Spï¿œnola, Ph.D.
Instituto Internacional en Conservaciï¿œn y Manejo de Vida Silvestre

Universidad Nacional
Apartado 1350-3000
Heredia
COSTA RICA
mspi...@una.ac.cr
mspin...@gmail.com

Telï¿œfono: (506) 2277-3598
Fax: (506) 2237-7036

[Richard Chandler]

I think the Jedediah was asking about models that allow for inference about species richness. See

Dorazio and Royle 2005 JASA 100:389
Dorazio et al 2006 Ecology 87:842
Royle and Dorazio (2008 Book) Chapter 12

Richard

_____________________________________
Richard Chandler, post-doc
USGS Patuxent Wildlife Research Center
301-497-5696

From:	Richard Schuster <ric.sc...@gmail.com>
To:	unma...@googlegroups.com
Date:	03/09/2012 10:39 AM
Subject:	Re: [unmarked] Multi-species models
Sent by:	unma...@googlegroups.com

---

De Wan also did it using occupancy models and in one of the PRESENCE tutorials and in their phidot.org forum I read about other people doing that too.

Cheers,
Richard S

Am 3/9/2012 07:35, schrieb Manuel Spínola:

I did it with occupancy models in unmarked but may be it was not appropiate, I don´t know. 

Best,

Manuel

2012/3/9 Richard Schuster <ric.sc...@gmail.com>

Hi Manuel,

As far as I know that's what some people are doing that use program PRESENCE (e.g. De Wan et al 2009 BiolCons).

Cheers,
Richard S

--
Manuel Spínola, Ph.D.
Instituto Internacional en Conservación y Manejo de Vida Silvestre

Universidad Nacional
Apartado 1350-3000
Heredia
COSTA RICA

mspi...@una.ac.cr
mspin...@gmail.com
Teléfono: (506) 2277-3598
Fax: (506) 2237-7036
Personal website: Lobito de río
Institutional website: ICOMVIS



--
Manuel Spínola, Ph.D.
Instituto Internacional en Conservación y Manejo de Vida Silvestre

Universidad Nacional
Apartado 1350-3000
Heredia
COSTA RICA

mspi...@una.ac.cr
mspin...@gmail.com
Teléfono: (506) 2277-3598
Fax: (506) 2237-7036
Personal website: Lobito de río
Institutional website: ICOMVIS

[Manuel Spínola]

What about using the new capture-recapture capability in unmarked to estimate species richness (using capture-recapture models)?

[Andy Royle]

yea you could do that if N[j] = number of species at a location "j"

the problem with this approach is that (1) the model doesn't preserve species identity across sample units and thus you cannot model species-level effects (the whole point of the Dorazio-Royle papers).
(2) related: you cannot model "heterogeneity" in encounter probability because unmarked does not handle model Mh in that context.

I actually think (2) would be easy to resolve if someone had a half-day to spend on it.....

andy



J. Andy Royle
Research Statistician

USGS Patuxent Wildlife Research Center

12100 Beech Forest Rd.
Laurel, MD 20708
http://profile.usgs.gov/professional/mypage.php?name=aroyle
andy_...@usgs.gov
phone: 301-497-5846
fax: 301-497-5545

Book: "Hierarchical Modeling and Inference in Ecology: The Analysis of Data from Populations, Metapopulations and Communities" by J. A. Royle and R.M. Dorazio.  

unmarked: A very useful R package for fitting certain hierarchical models using likelihood methods. Available from: http://cran.case.edu/web/packages/unmarked/index.html

A 5 hour "introduction to unmarked" Webinar can be found here: http://www.pwrc.usgs.gov/Royalvideo.cfm

[jaya]

Species richness models would be great, but honestly at this point
even a 2-species model would be very helpful. I can see how it would
be easy to include species B as a covariate to estimate occupancy (or
abundance) of species A, but what about the issue of imperfect
detectability for species B? Manuel, if you have code to do that, I
would really appreciate seeing how you did it!

Thanks everyone,

On Mar 10, 2:01 am, Andy Royle <aro...@usgs.gov> wrote:
> yea you could do that if N[j] = number of species at a location "j"
>
> the problem with this approach is that (1) the model doesn't preserve
> species identity across sample units and thus you cannot model
> species-level effects (the whole point of the Dorazio-Royle papers).
> (2) related: you cannot model "heterogeneity" in encounter probability
> because unmarked does not handle model Mh in that context.
>
> I actually think (2) would be easy to resolve if someone had a half-day to
> spend on it.....
>
> andy
>
> J. Andy Royle
> Research Statistician
> USGS Patuxent Wildlife Research Center
> 12100 Beech Forest Rd.
> Laurel, MD 20708http://profile.usgs.gov/professional/mypage.php?name=aroyle

> andy_ro...@usgs.gov

> phone: 301-497-5846
> fax: 301-497-5545
>
> Book: "Hierarchical Modeling and Inference in Ecology: The Analysis of
> Data from Populations, Metapopulations and Communities" by J. A. Royle and
> R.M. Dorazio.
>
> unmarked: A very useful R package for fitting certain hierarchical models
> using likelihood methods. Available from:http://cran.case.edu/web/packages/unmarked/index.html
>
> A 5 hour "introduction to unmarked" Webinar can be found here:http://www.pwrc.usgs.gov/Royalvideo.cfm
>
> From:
> Manuel Spínola <mspinol...@gmail.com>
> To:
> unma...@googlegroups.com
> Date:
> 03/09/2012 12:58 PM
> Subject:
> Re: [unmarked] Multi-species models
> Sent by:
> unma...@googlegroups.com
>

> What about using the new capture-recapture capability in unmarked to
> estimate species richness (using capture-recapture models)?
>
> Best,
>
> Manuel
>

> 2012/3/9 Richard Chandler <rchand...@usgs.gov>

> I think the Jedediah was asking about models that allow for inference
> about species richness. See
>
> Dorazio and Royle 2005 JASA 100:389
> Dorazio et al 2006 Ecology 87:842
> Royle and Dorazio (2008 Book) Chapter 12
>
> Richard
>
> _____________________________________
> Richard Chandler, post-doc
> USGS Patuxent Wildlife Research Center
> 301-497-5696
>
> From:
> Richard Schuster <ric.schus...@gmail.com>
> To:
> unma...@googlegroups.com
> Date:
> 03/09/2012 10:39 AM
> Subject:
> Re: [unmarked] Multi-species models
> Sent by:
> unma...@googlegroups.com
>

> De Wan also did it using occupancy models and in one of the PRESENCE
> tutorials and in their phidot.org forum I read about other people doing
> that too.
>
> Cheers,
> Richard S
>
> Am 3/9/2012 07:35, schrieb Manuel Spínola:
> I did it with occupancy models in unmarked but may be it was not
> appropiate, I don´t know.
>
> Best,
>
> Manuel
>

> 2012/3/9 Richard Schuster <ric.schus...@gmail.com>

> Hi Manuel,
>
> As far as I know that's what some people are doing that use program
> PRESENCE (e.g. De Wan et al 2009 BiolCons).
>
> Cheers,
> Richard S
>
> Am 3/9/2012 07:22, schrieb Manuel Spínola:
> Hi Jedediah and Richard,
>
> May be I am misunderstanding the question, but is it possible to use the
> species as a covariate (factor) in one of the unmarked models?
>
> Best,
>
> Manuel
>

> 2012/3/9 Richard Chandler <rchand...@usgs.gov>

> Hi Jedediah,
>
> These models won't be in unmarked for quite some time I'm afraid. But Dan
> Linden at MSU is working on it. And I might piece together a function for
> the workshop in April, but it won't be easy to add to unmarked. In the
> meantime, you can fit these models using BUGS or JAGS.
>
> Richard
> _____________________________________
> Richard Chandler, post-doc
> USGS Patuxent Wildlife Research Center
> 301-497-5696
>
> From:
> jaya <jedediah.bro...@gmail.com>
> To:
> unmarked <unma...@googlegroups.com>
> Date:
> 03/08/2012 09:28 PM
> Subject:
> [unmarked] Multi-species models
> Sent by:
> unma...@googlegroups.com
>

> Dear forum,
>
> I just wanted to check to see if any multi-species models have been
> implemented in Unmarked yet, or if anyone had any code they could
> share to help me with this.
>
> Beggars can't be choosy, but the Yamaura et al. (JAPPL 2011) multi-
> species RN model would be especially useful...
>
> Many thanks!
>
> --
> Manuel Spínola, Ph.D.
> Instituto Internacional en Conservación y Manejo de Vida Silvestre
> Universidad Nacional
> Apartado 1350-3000
> Heredia
> COSTA RICA

> mspin...@una.ac.cr
> mspinol...@gmail.com

> Teléfono: (506) 2277-3598
> Fax: (506) 2237-7036
> Personal website: Lobito de río
> Institutional website: ICOMVIS
>
> --
> Manuel Spínola, Ph.D.
> Instituto Internacional en Conservación y Manejo de Vida Silvestre
> Universidad Nacional
> Apartado 1350-3000
> Heredia
> COSTA RICA

> mspin...@una.ac.cr
> mspinol...@gmail.com

> Teléfono: (506) 2277-3598
> Fax: (506) 2237-7036
> Personal website: Lobito de río
> Institutional website: ICOMVIS
>
> --
> Manuel Spínola, Ph.D.
> Instituto Internacional en Conservación y Manejo de Vida Silvestre
> Universidad Nacional
> Apartado 1350-3000
> Heredia
> COSTA RICA

> mspin...@una.ac.cr
> mspinol...@gmail.com

[Manuel Spínola]

Thats right Andy.  I was only estimating species richness for sevral areas and I identify those areas as groups.  I cannot compare that model to other models and I use an heterogeneity model from MARK. 

What i did in MARK was to set the species in a column and the histry of detection and I identified the área for wich I wanted to estimate ricness.

Especies    HD            Group1  Group2

A              1001             1          0

B              1111             1          0 

C              0100             1          0

D              1010             1          0

A              0001             0          1

B              1000             0          1

Best,

Manuel

2012/3/9 Andy Royle <aro...@usgs.gov>

[Manuel Spínola]

Also, you can estimate species richness in unmarked using an occupancy model as is described in the Occupancy book.  You need to have a list of all possible species in the study area.

For a multiple-species model, what for me, is different than a species richnes model, I did (2 species as the most simple example):

The example is using occupancy

Data: mult  (I especified Species as a site covariate), You need to put all your sites repeated as many as the number of species that you want to model. 

Site  DH       Species

1      101          A

2       000         A

3       111         A

4       010         A

1      111          B

2      000          B

3      010          B

4      101          B

mod1 <- occu(~ Species ~ Species , mult)

mod2 <- occu(~ 1 ~ Species , mult)

In the mod1 I am asking for a detection probability different for each species and occupancy also different for each species.  May be I am wrong but that is what I did.

In mod2, detection probability is equal for the 2 species.  You can try other models.

Best,

Manuel

 

2012/3/10 jaya <jededia...@gmail.com>

mspi...@una.ac.cr
mspin...@gmail.com

[Hardin Waddle]

You should check out the methods of this paper.  It offers a method to model how the presence of Species A affects the occurrence of Species B when both are detected imperfectly.  It only works one way, so you have to assume that one species is dominant and is unaffected by the presence of the other.  This can be done easily in WinBUGS.

 

Waddle, J. H., R. M. Dorazio, S. C. Walls, K. G. Rice, J. Beauchamp, M. J. Schuman, and F. J. Mazzotti. 2010. A new parameterization for estimating co-occurrence of interacting species. Ecological Applications 20:1467-1475.

[Dan Linden]

The difference between the multi-species community model (to which
Richard Chandler is referring) and the DeWan et al (2009) approach is
the treatment of species as fixed or random effects.

DeWan et al (2009) used species as a fixed effect, and assumed that
the functional response to covariates was the same across species.
For instance, one of their models was psi(species + p/a). The
estimate of the fixed effect for perimeter/area ratio (p/a) was
potentially improved (i.e., greater precision) because it used the
response data from all the species combined. But this forced the
fixed effect for p/a to be the same for each species, albeit, with
species-specific intercepts. Now if they had included a species*p/a
interaction, then their model would no longer be pooling species
response data for the functional relationship of p/a and the results
would be the same as running separate regressions for each species.
In which case, there is no "improvement" in the modeling effort.

Using species as a random effect in a true multi-species community
model comes with it's own restrictions/assumptions, but it represents
a more flexible framework for modeling multiple species. This assumes
you have more than ~10 species, otherwise the random effects
distribution (or the prior) will not be adequately estimated. Like
Richard said, I'm trying to build on some code that Richard and Andy
(and Ian) have started to allow a multi-species model to be estimated
using maximum likelihood in unmarked. It is tricky though...


-----------------------
Daniel W. Linden, Ph.D.

Postdoctoral Research Associate
Dept. of Fisheries & Wildlife
Michigan State University
480 Wilson Road, Room 13
Natural Resources Bldg.
East Lansing, MI 48824

[Marina Xavier]

Hy everyone. I would like to know how to dealing with missing values at the matrix  of detection history on Unmarked?  In substitution of the dopt on presence? I couldn´t visit all the sites in the same way and, for this reason, I have some sites with no information about detectability and the same for some Observation-level covariates.

Because of this I failed to put the observation covariates on the model in R. 

I really appreciate if someone could help me.

Marina

[Jeffrey Royle]

hi Marina,
 most of the functions in unmarked should handle missing encounter history observations transparently -- you should just code those as NA in the encounter history matrix.

 If you have missing covariate values then you have some more trouble -- you should probably omit those observations from the data set, or else fill in with mean values, or some other method, and there has been discussion on this issue in previous posts.

regards
andy

--
You received this message because you are subscribed to the Google Groups "unmarked" group.
To unsubscribe from this group and stop receiving emails from it, send an email to unmarked+u...@googlegroups.com.
For more options, visit https://groups.google.com/groups/opt_out.