Time-varying trap covariates, multi-session data

[Murray Efford]

secr.fit() can allow trap covariates to vary by occasion (see ?timevaryingcov)*. This is true for both single- and multi-session data.

However, time-varying covariates are not compatible with the 'fastproximity' method that is the default for proximity detectors (this speeds up computation by collapsing occasions). secr.fit() tries to recognise when 'fastproximity' is inappropriate and switches it off (overriding any setting in details). However, in secr <= 5.4.3 it failed to recognise time-varying covariates in the multi-session case, and stuck with fastproximity as the default. That raises an error in secr.design.MS() called from secr.fit(), reported by Jaime Heiniger.

 

Old versions can be made to work by setting details=list(fastproximity = FALSE). The revised version 5.5.0 on GitHub now recognises time-varying covariates in the multi-session case (available on R-universe as explained in the Github readme file). Once recognised, that case switches to the slower but reliable fastproximity=FALSE.

 

* I'm not saying this is usually worth it, but hey, the software allows