High RB in secrdesign

[stefano palmero]

Hi all,

I am simulating a high-density population (about 7 individuals/100km2) using secrdesign over a polygon mask.

library(secrdesign)

densities <- c(0.005,0.007,0.009)
g0s <- c(0.03,0.04,0.06)
sigmasF <- c(700,900,1100)
sigmasM <- c(900,1000,1300)

scens <- make.scenarios(noccasions = c(4,8,12), nrepeats = 1,
                        D = densities,
                        sigma = sigmasF,
                        g0 = g0s,
                        groups = c("F","M"))

male <- scens$group == 'M'
scens$sigma[male] <- sigmasM

exfn <- function(x) {
  if (inherits(x, 'secr') & !is.null(x$fit)) {
    pred <- predict(x)
    pred[[1]]['D',]
  }
  else data.frame()
}

sims <- run.scenarios (100, scenarios = scens,
                        trapset = ls21, maskset = mask,
                        extractfn = exfn,
                        fit.args = list(model = sigma~h2,
                                        hcov = 'group'),
                        fit = TRUE, fit.function = 'secr.fit',
                        ncores = 10, byscenario = FALSE)

My results have high relative bias (>|0.10|) in 92820 cases out of 97200. In contrast, precision is often very high (RSE <0.20).

Why are my simulations struggling to be accurate in so many cases? Also, can I trust the results of the accurate ones?

Thank you very much and have a nice weekend!

[stefano palmero]

* Forgot to mention that I am using a polygon mask, and that could be problematic?

[Murray Efford]

Hi Stefano

I'd like to get to the bottom of this, but we really need the traps and mask to reproduce the problem. Perhaps send those offline if sensitive or too big to attach? Also, I'm not quite sure what you mean by a polygon mask. My approach would be to isolate one scenario that gives biased estimates.

Murray