Single season single species occupancy model

122 views
Skip to first unread message

satavisha banik

unread,
Jul 2, 2023, 8:51:36 AM7/2/23
to unma...@googlegroups.com
So I am new in occupancy model. I tried it in R unmarked. So I have 3 location and each location contains 81, 60, 56 effort days. Then sir suggest me to club the 3 locations as they are in same state and also club the effort days to 9.
I did it. And I have no observation covariates. I have only site covariates of 10 (elevation , slope, human population density, road density, and some from fragstats - forest_patchdensity, forest_gyrate, range_pd, range_gyrate and same to built-up ) 

So is that possible to create model by which I can interpret if the detection probability is varying by any of these site covariates..

Marc Kery

unread,
Jul 2, 2023, 11:57:24 AM7/2/23
to unma...@googlegroups.com
Dear Satavisha,

no, with a single site you can't separate the two parameter types in an occupancy: i.e., the probs of occupancy and detection. And more importantly, when you aggregate (that's what's meant by 'club', right ?) the original 3 sites to a single one, then it is impossible in principle to explain any variation among sites, because there is no variation among sites.

Even if you keep the original 3 sites separate, I would not recommend fitting any models with site covariates, because that would correspond to a regression with just 3 data points.

Sorry can't be more positive here.

Marc




From: unma...@googlegroups.com <unma...@googlegroups.com> on behalf of satavisha banik <satavis...@gmail.com>
Sent: Sunday, July 2, 2023 05:02
To: unma...@googlegroups.com <unma...@googlegroups.com>
Subject: [unmarked] Single season single species occupancy model
 
So I am new in occupancy model. I tried it in R unmarked. So I have 3 location and each location contains 81, 60, 56 effort days. Then sir suggest me to club the 3 locations as they are in same state and also club the effort days to 9.
I did it. And I have no observation covariates. I have only site covariates of 10 (elevation , slope, human population density, road density, and some from fragstats - forest_patchdensity, forest_gyrate, range_pd, range_gyrate and same to built-up ) 

So is that possible to create model by which I can interpret if the detection probability is varying by any of these site covariates..

--
*** Three hierarchical modeling email lists ***
(1) unmarked (this list): for questions specific to the R package unmarked
(2) SCR: for design and Bayesian or non-bayesian analysis of spatial capture-recapture
(3) HMecology: for everything else, especially material covered in the books by Royle & Dorazio (2008), Kéry & Schaub (2012), Kéry & Royle (2016, 2021) and Schaub & Kéry (2022)
---
You received this message because you are subscribed to the Google Groups "unmarked" group.
To unsubscribe from this group and stop receiving emails from it, send an email to unmarked+u...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/unmarked/CABZhgma-zoPFJ1hhWckfiTRX0zj9JOq6XTywwzL85h2wCa2kXw%40mail.gmail.com.

satavisha banik

unread,
Jul 3, 2023, 6:25:58 AM7/3/23
to unma...@googlegroups.com
Thanks for your concern, but I have to do occupancy for my dissertation then what should I do now!?

Marc Kery

unread,
Jul 3, 2023, 7:29:35 AM7/3/23
to unma...@googlegroups.com

Dear Satavisha,

 

Well, you can’t fit any models when you have essentially no data ☹

 

You could try to find some interesting data set from someone else ? or else you could study some interesting question about occupancy models using simulations ? perhaps something having to do with sample size/power calculation, where you could study how increasing the number or sites or visits affects the precision of the estimates of occupancy and/or detection.

 

Best regards  --- Marc

satavisha banik

unread,
Jul 3, 2023, 7:36:10 AM7/3/23
to unma...@googlegroups.com
Thanks for your help. 
If I add the observation covariates like cloud, wind and date ,time then is it possible?

Marc Kery

unread,
Jul 3, 2023, 7:41:32 AM7/3/23
to unma...@googlegroups.com

Yes, you do have replication along the detection axis. If you absolutely MUST fit an occupancy model to your data, come hell or high water, then you could keep your data set in the original format with 3 sites, keep the occupancy parameter at a mean (i.e., fit an intercept only), but add a couple of covariates into the detection part of the model.

 

Best regards  -- Marc

satavisha banik

unread,
Jul 8, 2023, 1:16:30 PM7/8/23
to unma...@googlegroups.com
Dear Marc,
 Thanks for all your help.
Could you please explain how to explain this numbers of intercept and estimates of forest_gyrate and pd ! Are they unit free or I have to rescale it?


Regards 

IMG_20230708_224236.jpg

Marc Kery

unread,
Jul 9, 2023, 4:09:09 AM7/9/23
to unma...@googlegroups.com
Dear Satavisha,

I am not sure I understand your question. Basically, you can interpret this occupancy model exactly like two logistic regressions. You have one logistic regression for occupancy probability and the other for detection probability.

In a logistic regression, you model a probability as a function of covariates not directly (this would be called with an "identity" link function), but instead with a logit link function. That is, your model looks like this:

logit(p_i) = log(p_i / (1 - p_i) ) = alpha + beta * cov_i + ....,

where p_i and cov_i are the probability and the covariate value, respectively, at site i.

So, yes, to understand how variation in a covariate is associated with variation in the probability, you usually want to use the inverse-link transformation. You can look this up in any book or treatment of logistic regression.

Best regards  --- Marc



Sent: Saturday, July 8, 2023 19:16
To: unma...@googlegroups.com <unma...@googlegroups.com>
Subject: Re: [unmarked] Single season single species occupancy model
 
Reply all
Reply to author
Forward
0 new messages