Estimating population size from density trend surfaces

[murray.efford]

This is an area where it's easy to go wrong: I would generally advise against it. Some of the issues are covered in Efford & Fewster (Oikos 2013). Remember:
1. It's really risky to extrapolate a trend surface beyond the range of your data (and this is often what you are doing if you use region.N).
2. The default link for density is 'log', not 'identity'.
3. The default predictions from predict.secr are at 'Base levels of covariates'; for x- and y-coordinates this means the centroid of the mask, as coordinates are standardized to mean zero inside the model. Use the 'savenew' argument in predict.secr to view the values for the default predictions.
4. predictDsurface() is often handy; try plotting the result and using spotHeight() to interrogate the surface.
Murray