Royle Nichols model vs standard model

questions concerning analysis/theory using program PRESENCE

Royle Nichols model vs standard model

Postby Justine33 » Thu Feb 12, 2015 1:27 pm

Hello everyone,

I have a general question related to the use of the Royle Nichols 2003 model vs the single season standard model. I have a set of 60 camera trap sites set at locations within the home range of my species of interest. I therefore aim to not assessing "true occupancy" but instead intensity of habitat use. To do this I originally intended to apply the standard single season occupancy model. However just in order to compare other model types I noticed that the Royle and Nichols 2003 model has much more support (based on AIC) for my data set, which probably implies that my data is subject to abundance induced heterogeneity between sites.

I was therefore wondering whether it is valid to use the Royle Nichols 2003 model in order to take into account abundance induced, with not an aim to assess abundance? And use it as a tool to look at the influence of covariates on intensity of habitat use? Or should I go back to the standard model.

Any advice would be much appreciated. Thank you.
Justine33
 
Posts: 10
Joined: Sun Jan 18, 2015 4:09 pm

Re: Royle Nichols model vs standard model

Postby jhines » Thu Feb 12, 2015 2:14 pm

The Royal-Nichols model is a different structure than the standard single-season model, so they cannot be compared using AIC. The RN model makes some strong assumptions about the distribution of animals, so I would suggest not using it unless you are comfortable with those assumptions and are interested in population size.
jhines
 
Posts: 599
Joined: Fri May 16, 2003 9:24 am
Location: Laurel, MD, USA

Re: Royle Nichols model vs standard model

Postby Justine33 » Thu Feb 12, 2015 4:42 pm

Thank you for the response. much appreciated.
Justine33
 
Posts: 10
Joined: Sun Jan 18, 2015 4:09 pm

Re: Royle Nichols model vs standard model

Postby jhines » Fri Feb 13, 2015 11:09 am

Justine (and Darryl),

My apologies for responding without verifying. I was confusing the Royal point count model with the Royal/Nichols heterogeniety model. Testing with expected values, the log-likelihood values for the Royal/Nichols abundance-induced heterogeneity model and the standard single-season model are virtually equal. So, it appears the AIC comparison is OK, and I think the answer to your original question is that it would be OK to use the RN model if you're comfortable with the assumptions. Does the RN model have more support than the standard model with covariates? Perhaps the covariates could explain the heterogeneity without going to the RN model.

Jim
jhines
 
Posts: 599
Joined: Fri May 16, 2003 9:24 am
Location: Laurel, MD, USA

Re: Royle Nichols model vs standard model

Postby Justine33 » Fri Feb 13, 2015 1:10 pm

Thank you Jim for clarifying this. Yes the RN model works better with covariates based on AIC, so ill investigate if the model assumptions fit the data set.

cheers.
Justine33
 
Posts: 10
Joined: Sun Jan 18, 2015 4:09 pm

Re: Royle Nichols model vs standard model

Postby pasqualotto » Wed Sep 26, 2018 11:07 am

Hi guys,

What about false positive occupancy models? Can I compare AIC of standard occupancy model (single season, single species) with Royle & Nichols and false positive at the same time?

I'm interested in explicitly test which of these parameterizations perform better for my data instead of assuming that one of them will be more appropriate. For detecting a hare species, I used two detection methods (camera-traps and footprints) with 205 sites.

Thank you!
pasqualotto
 
Posts: 3
Joined: Sun Aug 24, 2014 7:56 pm
Location: Ribeirão Preto, SP, Brazil

Re: Royle Nichols model vs standard model

Postby cooch » Wed Sep 26, 2018 12:24 pm

pasqualotto wrote:Hi guys,

What about false positive occupancy models? Can I compare AIC of standard occupancy model (single season, single species) with Royle & Nichols and false positive at the same time?

I'm interested in explicitly test which of these parameterizations perform better for my data instead of assuming that one of them will be more appropriate. For detecting a hare species, I used two detection methods (camera-traps and footprints) with 205 sites.

Thank you!


If the Royle & Nichols model has the same likelihood structure as the standard model, then, in theory, you can compare using AIC. If they don't, you can't. For AIC's to be comparable at all, the underlying likeliood structure needs to be the same. I'll leave it to you to determine the 'story' in this case, as part of your 'homework'. ;-)
cooch
 
Posts: 1628
Joined: Thu May 15, 2003 4:11 pm
Location: Cornell University

Re: Royle Nichols model vs standard model

Postby pasqualotto » Wed Sep 26, 2018 1:39 pm

cooch wrote:For AIC's to be comparable at all, the underlying likeliood structure needs to be the same. I'll leave it to you to determine the 'story' in this case, as part of your 'homework'. ;-)


I don't know if I really understood. The same likelihood structure means that the models under comparison need to have the same structure (number of parameters (k) and covariates) or the likelihood function implemented in each model?

Thank you so much!
pasqualotto
 
Posts: 3
Joined: Sun Aug 24, 2014 7:56 pm
Location: Ribeirão Preto, SP, Brazil


Return to analysis help

Who is online

Users browsing this forum: No registered users and 10 guests