I am using Bayesian zero-altered (hurdle) Poisson or negative binomial model using R2jags. Because the ACF of the Pearson residuals showed autocorrelation, I decided to apply autoregressive residual into the zero-altered model.
However, I don't know the autocorrelated error structure should be dealing with Pearson residual or simply "difference between fit value and observation".
Also, if it is the latter one, how can I calculate residuals after that?
