Random effects in single-season occupancy model

posts related to the RMark library, which may not be of general interest to users of 'classic' MARK

Re: Random effects in single-season occupancy model

Postby cooch » Thu Nov 01, 2018 7:16 pm

jlaake wrote:But isn't site = individual in occupancy? So if you are looking for a random effect in p, I can add it to RMark. Let me know.


Yup, I forgot this was a SS occupancy question (which is basically a closed abundance problem). My mistake. :oops:

Your options for heterogeneity in p are to fit logit-normal individual random effects, or finite mixtures. Both are available in MARK. So, if Jeff wants to make those accessible in RMark, you owe him -- beer works. ;-)
cooch
 
Posts: 1628
Joined: Thu May 15, 2003 4:11 pm
Location: Cornell University

Re: Random effects in single-season occupancy model

Postby darryl » Thu Nov 01, 2018 9:34 pm

Just to confirm, a model is desired with a random effect on detection and not on occupancy? The former is possible in MARK, but not the latter. And coming back to Evan's earlier point, why do you desperately need to include a random effect?

Cheers
Darryl
darryl
 
Posts: 495
Joined: Thu Jun 12, 2003 3:04 pm
Location: Dunedin, New Zealand

Re: Random effects in single-season occupancy model

Postby ms2018 » Mon Nov 05, 2018 12:07 pm

darryl wrote:Just to confirm, a model is desired with a random effect on detection and not on occupancy? The former is possible in MARK, but not the latter. And coming back to Evan's earlier point, why do you desperately need to include a random effect?



I was hoping to create a model with a random effect on occupancy. It's because of the way my study was designed... Without getting too much into the nitty gritty of it, I have 10 sites that are grouped into pairs, and within each pair, one site receives a "human use" treatment and one site is the "no human use" control. Then halfway through the summer this treatment switches to the other site within the pair. For the second summer, the site that ended summer 1 with a treatment, continues to receive the treatment for the first half of summer 2, and then the treatment/control switches again for the second half of summer 2. It's basically an ABBA format, and because of this, my detection histories for each species for each half of each summer are stacked on top of each other. So it looks like I have more sites than I do, which is why I need to include site as a random effect. Does this make sense? Or maybe I'm approaching this wrong...?
ms2018
 
Posts: 7
Joined: Sun Apr 22, 2018 6:05 pm

Re: Random effects in single-season occupancy model

Postby jhines » Mon Nov 05, 2018 12:49 pm

It seems to me that random effects are not needed here, since you know whether the treatment was applied for each survey. You can simply add a survey covariate indicating whether treatment was applied (humanUse=1), or not (humanUse=0). Then, model detection as a function of this covariate, possibly in conjunction with year.

If you've split each combination of site/season (where season is 1/2 of summer) as a site (row) of data, then you could model occupancy as a function of a (humanUse) site covariate. However, this means that your model says that occupancy is constant for the first 1/2 of summer, then changes in the interval between the last survey of the first 1/2 of summer and the first survey in the last 1/2 of summer. This could lead to heterogeneity in occupancy and biased estimates.
jhines
 
Posts: 599
Joined: Fri May 16, 2003 9:24 am
Location: Laurel, MD, USA

Re: Random effects in single-season occupancy model

Postby ms2018 » Tue Nov 06, 2018 6:05 pm

cooch wrote:
jlaake wrote:But isn't site = individual in occupancy? So if you are looking for a random effect in p, I can add it to RMark. Let me know.

Your options for heterogeneity in p are to fit logit-normal individual random effects, or finite mixtures. Both are available in MARK. So, if Jeff wants to make those accessible in RMark, you owe him -- beer works. ;-)


I am looking for a random effect in p, and if it can be added to RMark I would be eternally grateful!!
ms2018
 
Posts: 7
Joined: Sun Apr 22, 2018 6:05 pm

Previous

Return to RMark

Who is online

Users browsing this forum: No registered users and 16 guests

cron