Occupancy model for correlated sites

[jCeradini]

Hi Folks,

I'm working with a data set based on nested sampling. Each site contains many discrete "plots", and each plot was surveyed multiple times. We expect plots within sites to be correlated/dependent biologically. For now, I am interested in retaining plot level encounter histories rather than collapsing all information to the site level.

Are there models implemented in RMark to account for this correlated structure? In particular, I've been reading about the multi-method model (Nichols et al. 2008. J. Applied Ecology) and models for clustered sampling (Hines et al. 2010. Eco Apps). "Occupancy with correlated detections" in on the MarkModels list, although it is not currently implemented. Perhaps I am missing (or misinterpreting) another occupancy model on the list that would do the job?

I realize these models are available in PRESENCE, but would prefer to work in RMark if possible.

Thanks!
Joe

[jlaake]

Joe-

The definitive place to answer your question can be answered by looking at the list of models (Help/Data Types) in the MARK interface and see whether they have an asterisk to show they are supported in RMark. I'm not sure when I last updated MarkModels.pdf.

Once you know what model in MARK that you want to fit. If it is not currently supported by RMark then you are welcome to ask me to put it in. I'll evaluate how much work is involved to do so. Some of the Occupancy models are so different than traditional capture-recapture models that they require specific code which is not always easy to write.

I am more likely to say yes, if you set up an example model in the MARK interface, do a save structure (yellow funky arrow) and send me the save structure input file which is what I have to create in RMark.

--jeff

[jCeradini]

Thanks Jeff.

Unfortunately, I cannot get Help > Data Types to open (MARK 8.1). The MARK GUI opens fine but when I click on Help > Data Types I get "Cannot find the C:\Program Files(x86)\MARK\mrk5971z.tmp file. Do you want to create a new file?", Yes, No, or Cancel. I've uninstalled and reinstalled MARK twice in the last 30 mins, but no luck. I realize this is not your purview, Jeff, so I will post the problem more generally or ask Gary.

Joe

[jlaake]

Joe-

Make sure you are running the Mark interface in working directory where you have write permission. My guess is that is the problem you are having particularly if it is newly installed.

--jeff

[jCeradini]

I do have admin privileges on my work computer, but ya never know, so I tried MARK 8.1 on 2 different person computers and got the same error for Help > Data Types.
Does it open for you on 8.1?

Thanks again.

[jCeradini]

...2 different *personal computers...

[darryl]

Joe,
If it's the R rather than the MARK aspect for why you don't want to use PRESENCE, there is a beta RPresence package available from the PRESENCE website that allows you to fit single-season multi-scale models.
Cheers
Darryl

[jlaake]

Joe-

I downloaded 8.1 of MARK and as I expected when I ran from a directory outside of the Program Files(x86)/Mark directory it worked fine and when I was in that directory it fails and gives the error you saw. I have Admin privilege as well but I knew it would fail anyhow if it is open in the default directory.

To be specific, copy the examples directory of MARK to your desktop or someplace other than where it is by default. Open one of the .dbf files and then try Help/Data Types. It will work. None of the help files will work from inside the Program Files directory. Thank Microsoft and Windows. Next time please try what someone is telling you before replying.

--jeff

[jCeradini]

Thanks Darryl! I will check out RPresence as well.

You're right, Jeff. It worked when I followed your instructions in your last post. I would have happily tried what you suggested earlier but, I now realize, I did not understand what you meant (i.e., I did not understand that leaving the MARK folder in its default location meant that I necessarily could not have write permissions). I attempted to be thorough by trying the same process on 3 different computers to see if it was a privileges issue specific to my work computer, but clearly I was going down the wrong road.

Well, after a bit of detour on my end, looks like the occupancy models for correlated sites or detections are not yet implemented in RMark. Jeff, I'll need to look into some of the differences in the MARK models before making any specific appeals for an RMark addition. Specifically, models 143, 151-154 all look useful for data from nested sampling, but I need to suss out the differences.

Thanks again.
Joe

[jCeradini]

For the sake of a complete thread, RMark does support a multi-scale occupancy model (123. #*MultScalOcc: Multi-scale occupancy estimation) which, as Darryl suggested, can be used for nested data. There are a lot of models on that list, makes your eyes go blurry

Joe