Problem in running OpenCR

[Ujjwal]

Hi all

I am facing issues with OpenCR models with my 7 years of multisession data (leopard). I am unable to run the spatial open models. After running the models it throws an error

Error: cannot allocate vector of size 1.4 Gb

 I tried with 32GB RAM  with 32 Core system The error is the same cannot allocate vector of size only the memory size differs. The models are listed below

m1 <- openCR.fit(CH,mask=msk, type = 'CJSsecr',model = list(lambda~t, phi~t))

m.s<- openCR.fit(CH,mask=msk, type = 'CJSsecr',model = list(lambda~t, phi~sex))

Could you have any solution to deal this?

regards:

Ujjwal

[Murray Efford]

Hi Ujjwal

You imply that the model fits and then falls over - that is odd. lambda is not a parameter in the CJSsecr model, and maybe the problem has something to do with that. Otherwise, I think you're going to have to trouble shoot this yourself by trying simpler models or compacting the data.

- try non-spatial models first

- try constant models first

- check the size of your habitat mask (around 1000 points should be enough for a start)

- do not use daily 'occasions' within each year

Note also that CJSsecr is unsafe, and ultimately you will want to use JSSAsecrf or similar.

Murray

--
You received this message because you are subscribed to the Google Groups "secr" group.
To unsubscribe from this group and stop receiving emails from it, send an email to secrgroup+...@googlegroups.com.
For more options, visit https://groups.google.com/d/optout.

[Ujjwal]

Thanks, Murray

I have tried after collapsing occasion as Count (reduce function of secr) and It works. Thanks for giving clue in dealing the memory error.

However, I am having issues of parameter non-identifiability (The Hessian rank is 0 and Eigen Value is not calculated )basically with  the male tiger capture history, The female tiger data works well. I assumed the movement between the primary session is normal. The sub-adult males dispersed in next primary period. Hence their Home range centre will change. I will  be thankful if you could give some idea to deal with non-identifiability in this scenario. 

regards:

Ujjwal

[Murray Efford]

I don't have any particular insight on this. Models with dispersal between sessions often seem hard to fit and possibly require better data than you have.