MCgof unable to handle cases when sites were not sampled on first occasion

[fbled]

Hello,

I have been trying to use MCgof on a model fitted in a situation where some sites where not sampled on the first occasion, and this seems to cause some issues with mismatch dimensions in detectpar0().

In MCgof, call goes down MCgof > MCgof.secr > simfxiAC > detectpar0

In detectpar0(), I have tracked the issue down to the following section :

  arglist <- c(list(x = object$design0$PIA, drop = FALSE), 
    indlist)
  selected <- do.call("[", arglist)
  out <- as.data.frame(realparval0[selected, , drop = FALSE])
  dd <- lapply(dim(selected), function(x) 1:x)
  names(dd) <- c("session", "animal", "occasion", "trap", 
    "class")
  id <- do.call(expand.grid, dd)[, sapply(indlist, is.logical), 
    drop = F]
  cbind(out, id)

If a site i was not in use during the first session, object$design0$PIA[1,1,1,i,] is set to 0, and it causes this site to get dropped in out, leading to a mismatch between out and id. This is how far I was able to track the issue down.

[MurrayEfford]

I see the problem. It can be reproduced with:

library(secr)

ch <- subset(captdata, traps=10:76)
usage(traps(ch)) <- matrix(1,nrow=67,ncol=5)
usage(traps(ch))[1,1] <- 0
fit <- secr.fit(ch)

secr:::detectpar0(fit)
Error in data.frame(..., check.names = FALSE) :
arguments imply differing number of rows: 0, 1

Thanks for pointing this out. I'll work on a fix, but it may not be immediate. Meantime, a workaround is to reorder the traps:

reordertraps <- function (ch, neworder = c(2:dim(ch)[3], 1)) {
traps(ch)[,] <- traps(ch)[neworder,]
usage(traps(ch))[,] <- usage(traps(ch))[neworder,]
ch[,,] <- ch[,,neworder]
ch
}
ch2 <- reordertraps(ch)
fit2 <- secr.fit(ch2)

secr:::detectpar0(fit2)
g0 sigma session
1 0.2507396 29.0295 1

[fbled]

Thanks for the reply, and apologies for forgetting the repex! And thanks for the fixes (both temporary and upcoming!)