murrayefford / opencr Goto Github PK

openCR ignores overdispersion. It is worth investigating how it could be included (cf Fletcher; Bischof et al.)

It should be straightforward to add capability for modelling selective settlement to the present algorithm for truncation-and-normalisation at the boundary. It might require new structures for modelling functions of mask covariates, as used in secr for D, noneuc.

A new function 'modelAverage' is under development.

It's actually a generic, with methods modelAverage.openCR, modelAverage.openCRlist. It exists in openCR but is not exported, pending tests and extensions to cover the full range of models. It may be called at present with openCR:::modelAverage(). Arguments mostly follow secr::model.average. The default criterion is 'AIC' not 'AICc' as in 'secr'.

To do:

• check modelAverage.openCRlist works with complex newdata, including stratified analysis