Unsure of proper model to use

Forum for discussion of general questions related to study design and/or analysis of existing data - software neutral.

Unsure of proper model to use

Postby sarah1795 » Wed Apr 24, 2019 4:13 pm

My apologies for this being such a basic question but I greatly appreciate any insight, guidance, or suggestions.

I sampled 2 sites over a 6 month period and have my time interval set to days. I coded my capture histories are by the individual animal and I used dot notation '.' for the times when I was the opposite site from the individual.

Site 1
Animal A ...010

Site 2
Animal B 010...

I'm interested in estimating abundance of animals at these two sites, using RMark, and I'm having trouble finding the best model to do this in that allows for dot notation - Jolly, Burnham, Robust, etc give errors that ch can only contain '01'. I am very new to modeling and have been reading through the Gentle Introduction, sifting through literature, and emailing most of my department... from this I'm thinking perhaps POPAN or JS model but I'm not if my data is correct for them. I'm not sure how to define N properly within RMark to get POPAN to run without error. I can run CJS without R spitting out an error but I'm not sure how to go from there to abundance.

A brief-ish description of my study for background:
Live trapped and ear tagged 11 species (4 species with over 10 individuals) of small mammals at 2 sites ~100 miles apart. Trapping grid was 200 Sherman traps 10m apart, roughly square. Sampled each site 6 consecutive days per month May-October. Recorded species, ear tag #, body length, weight, age class (adult/juv), sex (except shrews), and tick burden.
sarah1795
 
Posts: 1
Joined: Wed Feb 20, 2019 10:27 am

Re: Unsure of proper model to use

Postby murray.efford » Thu Apr 25, 2019 11:01 pm

A more fundamental problem is that population size N is not a meaningful parameter unless each of your grids spans a natural habitat island. I suggest you look at spatially explicit methods, for which density D is the abundance parameter. Spatially explicit methods routinely allow for spatially incomplete sampling on a particular occasion. Seems you have sparse data that you intend to pool across species; this can be done for spatially explicit analyses as for nonspatial analyses, with the same risks.
murray.efford
 
Posts: 686
Joined: Mon Sep 29, 2008 7:11 pm
Location: Dunedin, New Zealand

Re: Unsure of proper model to use

Postby ehileman » Sat May 04, 2019 10:04 am

My apologies for this being such a basic question but I greatly appreciate any insight, guidance, or suggestions.

I sampled 2 sites over a 6 month period and have my time interval set to days. I coded my capture histories are by the individual animal and I used dot notation '.' for the times when I was the opposite site from the individual.

Site 1
Animal A ...010

Site 2
Animal B 010...

I'm interested in estimating abundance of animals at these two sites, using RMark, and I'm having trouble finding the best model to do this in that allows for dot notation - Jolly, Burnham, Robust, etc give errors that ch can only contain '01'. I am very new to modeling and have been reading through the Gentle Introduction, sifting through literature, and emailing most of my department... from this I'm thinking perhaps POPAN or JS model but I'm not if my data is correct for them. I'm not sure how to define N properly within RMark to get POPAN to run without error. I can run CJS without R spitting out an error but I'm not sure how to go from there to abundance.

A brief-ish description of my study for background:
Live trapped and ear tagged 11 species (4 species with over 10 individuals) of small mammals at 2 sites ~100 miles apart. Trapping grid was 200 Sherman traps 10m apart, roughly square. Sampled each site 6 consecutive days per month May-October. Recorded species, ear tag #, body length, weight, age class (adult/juv), sex (except shrews), and tick burden.

Top


Hi Sarah,

To use dot notation in your capture histories, you'll need to select a model that uses conditional likelihood. Jolly-Seber and Popan models use full-likelihood and therefore do not permit dots. If you can meet the assumption of closure, the Huggins closed model within a capture-recapture or SECR framework may be appropriate. Six days should be sufficiently short to assume closure. To estimate abundance for each species, site, and month simply code each species, month, and site as separate grouping variables. It sounds like your data may be a bit sparse though, so stratifying your data this way may not be possible. You may need to pool some groups.

Alternatively, you could use the open population CJS model (another conditional likelihood model) and the Horvitz-Thompson estimator to estimate abundance if the assumption of closure cannot be met (See McDonald et al. 2005). Hope this helps!

Eric

    McDonald, T. L., S. C. Amstrup, E. V. Regehr, and B. F. J. Manly. 2005. Examples. Pages 196-265 in S. C. Amstrup, T. L. McDonald, andB. F. J. Manly, editors. Handbook of Capture-Recapture Analysis. Princeton University Press.
ehileman
 
Posts: 51
Joined: Sat Nov 26, 2011 6:40 pm
Location: West Virginia University

Re: Unsure of proper model to use

Postby jlaake » Sun May 05, 2019 8:45 am

All of the robust design Huggins models like RdHuggins are supported in RMark.
jlaake
 
Posts: 1417
Joined: Fri May 12, 2006 12:50 pm
Location: Escondido, CA

Re: Unsure of proper model to use

Postby murray.efford » Mon May 06, 2019 3:23 am

There is no need to use conditional likelihood (Huggins) approaches to allow for spatially incomplete sampling in SECR models, but the option is there if you want to include individual covariates. I stick to my original point: it is almost certainly a mistake to use non-spatial models with these data. It's primarily a question not of software but of what parameter you are trying to measure, as explained before. I suspect the CJS route is only of historical interest, even in the non-spatial realm - it's waiting for someone to do a proper comparison.

If you go down the SECR path you have a choice between treating each month as a separate closed population (R package secr), running a multi-session closed-population analysis (also in secr) or an open population robust design SECR model (R package openCR, or Bayesian options). The multi-session and open-population options allow sharing of detection parameters across months; the extra power may be important.
murray.efford
 
Posts: 686
Joined: Mon Sep 29, 2008 7:11 pm
Location: Dunedin, New Zealand


Return to analysis & design questions

Who is online

Users browsing this forum: No registered users and 13 guests

cron