glimmix dist=lognormal vs autoreg

Question

I am working with time series data in SAS proc glimmix. When I change the dist=lognormal, the models fit much better. My question is, would this be an acceptable approach vs autoreg? From my research there does not seem to be an option for autoreg dist=lognormal. I would be happy to provide any of the outputs. Appreciate the help.

Answer 1

The choice of distribution is related to how your response variable is distributed. I would plot the response variables' histogram and the relationships with the regressors before modelling.