GOF - closed population models

questions concerning analysis/theory using program MARK

GOF - closed population models

Postby Chris Jones » Tue May 10, 2005 6:07 pm

What is the most appropriate method for assessing the GOF of closed population models in MARK?
Chris Jones
 
Posts: 6
Joined: Tue Oct 19, 2004 11:16 pm
Location: New Zealand

RE: GOF - closed population models

Postby bmitchel » Thu May 12, 2005 1:32 pm

Chris -

As you have no doubt discovered, there is no way to test GOF of closed capture models in MARK. In addition, most of the obvious bootstrap methods (e.g. based on the deviance statistic reported by MARK) are poorly studied, and are presumably biased.

I recently read a paper by MacKenzie and Bailey (2004, Assessing the fit of site-occupancy models, Journal of Agricultural, Biological, and Environmental Statistics 9(3):300-318) that presents a bootstrap approach based on a revised chi-square statistic (i.e. not the chi-square statistic in MARK output). The authors assert that their method should work for any mark-recapture models, with or without covariates.

I am not aware of any simulation studies that apply this approach to mark-recapture data to test the authors' assertion, but I found the occupancy results convincing enough that I have been using this approach to test GOF of closed-capture removal models with covariates.

Brian
bmitchel
 
Posts: 28
Joined: Thu Dec 09, 2004 9:57 am

Re: RE: GOF - closed population models

Postby darryl » Sun May 15, 2005 7:01 pm

[quote="bmitchel"]

I am not aware of any simulation studies that apply this approach to mark-recapture data to test the authors' assertion, but I found the occupancy results convincing enough that I have been using this approach to test GOF of closed-capture removal models with covariates.
[/quote]

Hi All
I have played around with this technique Brian cites for CJS-type (ie recaptures only) data and it's seems to work ok. It's basically an off-shoot/generalization of Gary et al's parametric bootstrap test that's already in MARK, except using Pearson's chi-square statistic instead of the deviance. You don't get the same negative bias in c-hat for CJS-type data that you do with the current parametric bootstrap, but estimates can be highly variable. I don't know how it compares to the median c-hat test or some of the latest GOF tests that have been developed. Nor have a I tried it out on other data types. There's a good research project if anyone's interested.. :wink:
Cheers
Darryl
darryl
 
Posts: 498
Joined: Thu Jun 12, 2003 3:04 pm
Location: Dunedin, New Zealand


Return to analysis help

Who is online

Users browsing this forum: Google Adsense [Bot] and 1 guest