I have a question regarding GAM models using count data and the Poisson family. Here the data set I am using. In the photos: model summary and plots generated using gam.check from the R package "mgcv". My questions are about some of the plots. I am finding a bit of difficulty reading them considering I am using count data:
- the "Resids vs linear pred." plot: does that downward trend suggest that the model needs improvement?
- the "Response vs fitted" plot: is this ok for count data? Thanks so much, and I will be happy to elaborate if the information I am providing is not enough