Data where individuals are temporarily unavailable

questions concerning analysis/theory using program PRESENCE

Data where individuals are temporarily unavailable

Postby vegemite » Fri Jan 31, 2020 2:47 pm

Hello,

Thanks for visiting my question. I have spent the past few weeks reading up on occupancy analysis, and looking through this forum has been very helpful.

I have presence/absence data for a single species at 63 sites recorded over 20 weekly surveys. The 20-week survey period includes a pre-hibernation, hibernation and post-hibernation stage – so generally speaking, ‘occupied sites’ tended to be used during the first few weeks, unused during mid-season, and then used again during the last few weeks. The individuals’ temporary unavailability during hibernation possibly violates the closure assumption but that’s another issue I’m trying to get my head around.

I am interested in assessing whether covariates had different impacts on occupancy/detection during the different stages in the survey period. For example, I’d like to look into whether a drop in site use was associated with temperature change, or whether a continuation in site use (through the supposed hibernation period) was associated with supplementary feeding.

To look at these different stages of the hibernation period, would it be best to just split the data into three separate datasets/projects (pre-hibernation, hibernation, and post-hibernation)?

My second question relates to rather high c hat values (>40) and tackling over-dispersion. Would a potential solution be pooling surveys so that 2 weeks (or more) of data become 1 survey? e.g. detection history 01 would become 1.

Thanks for your time.
vegemite
 
Posts: 12
Joined: Fri Jan 31, 2020 10:57 am
Location: UK

Re: Data where individuals are temporarily unavailable

Postby darryl » Sun Feb 02, 2020 4:43 pm

From your description it would sound reasonable for you to split into 3 periods and use a multi-season analysis so you can assess whether covariates have the same or different effects in each period.

It depends on the reason for a high c-hat value as to what you should do. A value of 40 is really extreme and would indicate the model is really doing a poor job of modelling the data, or you can also get high values like that if you have sparse data or data with lots of missing values. Hard to give any suggestions without knowing more.

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

Re: Data where individuals are temporarily unavailable

Postby vegemite » Fri Feb 07, 2020 3:04 pm

darryl wrote:From your description it would sound reasonable for you to split into 3 periods and use a multi-season analysis so you can assess whether covariates have the same or different effects in each period.

It depends on the reason for a high c-hat value as to what you should do. A value of 40 is really extreme and would indicate the model is really doing a poor job of modelling the data, or you can also get high values like that if you have sparse data or data with lots of missing values. Hard to give any suggestions without knowing more.

Cheers
Darryl


Hi Darryl, thank you for your response. I think my problem with the c-hat was relating to rather sparse data. The chi square values were extremely high for some detection histories. However, when I split the data into 3 periods and perform 3 single-season analyses, the c-hat value comes right down. So chunking the data seems to be the best way forward.

A quick follow-up Q - when using multi-season analysis, must the number of secondary sample occasions per each season be an equal number, e.g. 5, 5, 5, 5? In an ideal world mine would be split into 5, 10, 5 (since that corresponds with the hibernation period timings), but when I try to enter this into PRESENCE, or even edit an existing dataset in PRESENCE, it freezes up. Actually, it looks like from your 2003 Ecology paper, that if I wanted to split it into my ideal 5, 10, 5, I'd have to do 10, 10, 10 with missing observations.
vegemite
 
Posts: 12
Joined: Fri Jan 31, 2020 10:57 am
Location: UK

Re: Data where individuals are temporarily unavailable

Postby jhines » Fri Feb 07, 2020 4:11 pm

You can change the number of secondary surveys to 5,10,5 if you go to the View/Data menu and change the box at the top labelled, '#srvy/seasn'. After making the change, save the file (File/Save as) in the project folder, then exit Presence and restart Presence.

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

Re: Data where individuals are temporarily unavailable

Postby vegemite » Tue Feb 11, 2020 7:02 am

jhines wrote:You can change the number of secondary surveys to 5,10,5 if you go to the View/Data menu and change the box at the top labelled, '#srvy/seasn'. After making the change, save the file (File/Save as) in the project folder, then exit Presence and restart Presence.

Jim


Thank you for the reply. I am facing some sort of technical issue as Presence crashes when I start typing '5' into the #survey/season box. I'll try installing the slightly newer version!
vegemite
 
Posts: 12
Joined: Fri Jan 31, 2020 10:57 am
Location: UK

Re: Data where individuals are temporarily unavailable

Postby jhines » Tue Feb 11, 2020 10:05 am

I haven't encountered that problem before. If the new version doesn't work, let me know. Perhaps we could resolve it via a skype call.

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

Re: Data where individuals are temporarily unavailable

Postby vegemite » Wed Feb 12, 2020 1:25 pm

jhines wrote:I haven't encountered that problem before. If the new version doesn't work, let me know. Perhaps we could resolve it via a skype call.

Jim


Hi Jim,

Installing the newer version didn't make a difference. However, I did find a way to solve this problem. The issue seemed to actually be when I typed in the comma following '5'. So instead I typed '5 10 5' in with no commas, and went back and added the commas afterwards. This worked fine! Perhaps this is me making a silly mistake and that's how you're supposed to enter it into the #srvey/seasn box anyway?! In any case, thanks for your offer of help.

Best wishes,

Abi
vegemite
 
Posts: 12
Joined: Fri Jan 31, 2020 10:57 am
Location: UK

Re: Data where individuals are temporarily unavailable

Postby jhines » Wed Feb 12, 2020 1:50 pm

I'm glad you got it to work. Just make sure to check the number of secondary periods listed in the output files that they were read in correctly.

Looking at my code, I don't see why that would work as it looks for comma's to separate the values. Are you using a computer with a different (from English) language version of Windows?

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

Re: Data where individuals are temporarily unavailable

Postby vegemite » Wed Feb 12, 2020 3:00 pm

jhines wrote:I'm glad you got it to work. Just make sure to check the number of secondary periods listed in the output files that they were read in correctly.

Looking at my code, I don't see why that would work as it looks for comma's to separate the values. Are you using a computer with a different (from English) language version of Windows?

Jim


Hi Jim,

I am using Windows 10 (English GB), 64-bit, with keyboard language set to English UK. I carried out a multi-season analysis a few weeks ago on this laptop and had no problem entering commas then, IIRC. Not sure what has changed in those few weeks but perhaps it is just a problem on my end rather than with Presence itself.

Abi
vegemite
 
Posts: 12
Joined: Fri Jan 31, 2020 10:57 am
Location: UK


Return to analysis help

Who is online

Users browsing this forum: No registered users and 5 guests