multi-season heterogeneity models?

questions concerning analysis/theory using program PRESENCE

multi-season heterogeneity models?

Postby dtempel » Fri Apr 01, 2011 6:27 pm

Hi everyone. I noticed that a single-season heterogeneity (i.e. mixtures) model is available in PRESENCE, but I don't see the option for a multi-season heterogeneity model. Is there a reason for this (e.g. the method hasn't been fully developed yet)?

I'm interested in using mixtures to model detection probability in a multi-season analysis. Thanks.
dtempel
 
Posts: 27
Joined: Mon Jul 02, 2007 4:55 pm
Location: Tucson, AZ

Re: multi-season heterogeneity models?

Postby jhines » Tue Apr 05, 2011 2:55 pm

Hi,

I don't know of anyone who has asked about this before, so I doubt there is any software to handle it. Are you think of using it for a species-richness type of application? Or do you have some sites which you believe have different detection probabilities which you don't have covariate information on?

Jim
jhines
 
Posts: 632
Joined: Fri May 16, 2003 9:24 am
Location: Laurel, MD, USA

Re: multi-season heterogeneity models?

Postby dtempel » Wed Apr 06, 2011 5:30 pm

It's the latter case. We have some sites that I believe have different detection probabilities but I lack covariate data that might explain why they're different. I'm going to use reproductive status during each year as a site-specific covariate, but I suspect there's some other reason for the variation. It seems like we have some sites where the birds are consistently detected on nearly every survey, but other sites where we consistently have only one (or maybe two) detections each year. This is a spotted owl data set, and the sites are owl territories that have been surveyed every breeding season for a number of years.

My intuition tells me that the birds at those sites may be moving around a lot, so they may not be "available" for detection on every survey. I know that violates a model assumption, which would lead me to interpret occupancy in this case as the proportion of sites that are "used" each year.

It looks like Program MARK has the option to run multi-season heterogeneity models, but I also want to run some models with survey-specific covariates (e.g., detection probability changes after the initial detection at a site during a given year). It's not clear to me how I would input survey-specific data in MARK. Site-specific covariates are simple enough, but I haven't figured out survey-specific covariates.

Thanks!
dtempel
 
Posts: 27
Joined: Mon Jul 02, 2007 4:55 pm
Location: Tucson, AZ

Re: multi-season heterogeneity models?

Postby jlaake » Wed Apr 06, 2011 5:56 pm

In MARK, create a separate individual covariate for each survey. For example, if you had precip as 0/1 for each of 3 surveys then you might create precip1,precip2,precip3 as the variable names and the values would be 0 or 1 depending on whether there was precip on that survey for that site. In the design matrix, you could create a column for precip and in the rows for survey 1, you enter precip1, for survey 2, precip2 etc. Hope that helps.

--jeff
jlaake
 
Posts: 1480
Joined: Fri May 12, 2006 12:50 pm
Location: Escondido, CA

Re: multi-season heterogeneity models?

Postby dtempel » Wed Apr 06, 2011 6:25 pm

Yes, that makes sense. Unfortunately, we have 180 survey occasions in the data set (10 survey periods per year X 18 years), and not all sites were surveyed each survey period (i.e., we have missing data). I'd need 180 covariates for each site with dots for the survey periods when a site wasn't surveyed, correct?

Thanks,
Doug
dtempel
 
Posts: 27
Joined: Mon Jul 02, 2007 4:55 pm
Location: Tucson, AZ

Re: multi-season heterogeneity models?

Postby gwhite » Wed Apr 06, 2011 6:41 pm

Jeff:
You've making this problem way too hard by using individual covariates like precip1, precip2, etc. Just paste the values into the design matrix -- because what we're talking about is an environmental covariate that does not need to be an individual covariate.

Gary
gwhite
 
Posts: 340
Joined: Fri May 16, 2003 9:05 am

Re: multi-season heterogeneity models?

Postby darryl » Wed Apr 06, 2011 7:00 pm

Gary, that will only work if the same covariate value is going to apply to all sites at the same survey occasion (eg regional rainfall in the last 24 hrs), if it's not (eg site-specific rainfall in the last 24 hrs) the you will need to do has Jeff suggests. In Doug's application it sounds like he wants to create a trap-response covariate to allow the probability of redetection at a site to be different from first detection which is probably going to be of the latter type of covariate in general.
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: No registered users and 1 guest

cron