Hello, I have count data with a large number of zeros. To model this data using gams I want to try the Tweedie distribution instead. (The zero-inflated poisson willbe my next step) How do i decide on the value of the power 'p' parameter in the formula: family=Tweedie(p, link) ? For a glm, this can be found using the tweedie.profile function. But it doesnot work with the gam model formula where explantory variables are specified as s(x1) instead of x1. Will using a tweedie on discrete data affect my results? If my question has already been addressed before, a reference to the relevant information will be very much appreciated. Thanks, payal. [[alternative HTML version deleted]]