www.phidot.org

Hello.

I am trying to run density estimates with small mammal capture data over a number of years (great basin pocket mice and deer mice). I am running the analysis for each year separately with the trapping grid names as Session. My analysis works well for all years except for 2020 and 2022. For some reason that I can't determine, the data for these years both turn bakc an initialsigma value of 0m/n, and I am unable to run the model with this. Does anyone have any suggestions regarding how I should address this or what issues I should look for? The message I get back from R is below.

> initialsigma <- RPSV(sm2020, CC = TRUE)[[1]]
> cat("Quick and biased estimate of sigma =", initialsigma, "m/n")
Quick and biased estimate of sigma = 0 m/n
> ## Simple SECR model fit
> setNumThreads(7) #number of cores to use
[1] 7
> fit <- secr.fit(sm2020, buffer = 4*initialsigma, trace = FALSE)
Warning messages:
1: In make.mask(tr, ...) : no points in mask
2: In make.mask(tr, ...) : no points in mask
3: In make.mask(tr, ...) : no points in mask
4: In make.mask(tr, ...) : no points in mask
5: In autoini(capthist = ch, mask = msk, binomN = tempbinomN, adjustg0 = details$binomN[1] == :
'dbar' close to zero; using detector spacing instead
6: In makeStart(start, parindx, capthist, mask, detectfn, link, details, :
'secr.fit' failed because initial values not found (data sparse?); specify transformed values in 'start'
> fit
$call
secr.fit(capthist = sm2020, buffer = 4 * initialsigma, trace = FALSE)

$fit
NULL

It looks as if you have no 'spatial recaptures' on the first grid in those years i.e. any recaptures were at the same site as the initial capture. So basing 'buffer' on that grid is a mistake.

There is no need to use a year-specific buffer width, and I think it would be clearer for everyone if you settled on one number (and the habitat mask that implies) large enough for all years and grids. The RPSV call just gives you an idea what that number should be.

Thanks for the response! I appreciate the help.

Unfortunately, I'm pretty new to these types of models, as I'm an undergrad working on my Honours thesis. So, I'm hoping I can clarify some things with you. I've been using code from the tutorial and tweaking it here and there to apply it to my data as needed. I understand what you're saying but am unsure how to address it in my code. Should I be skipping the RPSV command altogether and using different code or using the initialsigma value from my year with the widest buffer to apply to all years?

Basically the second option: choose an adequate buffer width and stick to it. And check out https://www.otago.ac.nz/density/pdfs/secr-habitatmasks.pdf.