first question here. I'm constructing GAMs for multispecies avian survey data, and running into some odd results in the calculation of R.sq in the summary() command. For one of my species, when i use the negative binomial response distribution, i get a large negative value (R.sq = -4.54), but when i use the Tweedie distribution, I get an R.sq of 0.4. Deviance explained in either case are realistic values (30-40%), and ensuing term selection results in identical models for either distribution. What gives here? Is R.sq calculation different for Tweedie and negative binomial? If anything i would have thought Tweedie's adjusted R.sq would be lower since it has an extra parameter to be estimated.
If it helps, I have ~950 obs with ~80% zeros.