GAMs: predictions with the number of live traps as an offset variable
I built a Generalized Additive Model (GAM) using a negative binomial distribution (my response variable represents an animal count). In the GAM, I included the log(trapping effort) as an offset variable to account for variations in the sampling effort across trapping sites. Trapping effort was calculated as the total number of live traps per night. Is there a way to fit the GAM at each raster pixel defining a landscape when the trapping effort in the pixel is unknown (some landscape areas do not cover the trapping sites)? In particular, I would like to use the GAM in order to build a predictive map of abundance.