Large errors from predict() and occuPEN help

[Bethany Smith]

Hi everyone,

I'm really hoping somebody can please help as I'm quite new to occupancy modelling and unmarked. I have camera trapping data for 29 sites over 4 months and have been running single species models quite easily. However, for one species when using predict to get predicted means and 95% CIs I get huge confidence intervals essentially from 0 to 1 and the predicted mean occupancy is 1. I have 378 detections of this species, which grouped into weekly visits results in 125 presences vs 133 absences. The naive occupancy is 0.86 and the null model occupancy estimate is 0.90, so they're pretty much all over my study area. Plotting the predicted outputs looks like this:

I think this is a boundary estimate issue due to all my sites at the low habitat type being occupied so the estimate is 1. I looked into using penalized likelihood methods but I'm not sure I quite understand how to implement them.

I first tried the MPLE method but when I try to run the computeMPLElambda function I get the following error:

Error in if (any(y < 0 | y > 1)) stop("y values must be 0 <= y <= 1") :
  missing value where TRUE/FALSE needed

Is this because I have NAs in the data where cameras were not functioning? If so, is there a way around this?

As this wasn't working for me I also tried the occuPEN_CV method with Ridge and Bayes. I seem to be able to run these but I don't know how to interpret the outputs or get confidence intervals. Following the package documentation, I set up my model as such:

> testCV <- occuPEN_CV(~1 ~LGDpres, deer.data,
                       pen.type="Ridge", foldAssignments=rep(1:5,ceiling(nSites/5))[1:nSites])

Then, running 'testCV @estimates' I get NAs for everything bar the estimate:

Occupancy:
            Estimate SE  z P(>|z|)
(Intercept)    0.748 NA NA      NA
LGDpreslow     8.960 NA NA      NA

I have a feeling this is normal, so continued on to run:

> testCV <- nonparboot(testCV,B=100) 
> vcov(testCV,method="nonparboot")

And get the following:

I honestly have no idea if I've done this correctly, if it's useful, how to interpret the output, or how to get the back transformed occupancy estimates and confidence intervals.

If anybody can help me at all, even just pointing me in the right direction I would be extremely grateful. I'm not even sure the penalized methods are the correct option here, but I wasn't sure what else to try first.

Thanks in advance!

Beth