Recently, I was asked to check for serial correlation after doing a panel Poisson regression. I haven't seen such a test and in general, researchers (at least in the econom(etr)ics literature) don't report many diagnostics.

There is some discussion going on on this forum about the nature of residuals from a Poisson regression (see for example here, here and here), specifically about the questions 1) whether residuals should normally distributed, and 2) if not, what their theoretical distribution is. My understanding so far is: 1) no, and 2) it's complicated.

Am I correct in thinking that the distribution of residuals for panel Poisson analysis is also complicated, probably more so than for cross-section; and that because of that, there simply are not as many formal diagnostic tests as for linear models?

(My immediate question is about serial correlation in panel Poisson models, but I would be interested in a more general discussion for Poisson models, non-linear models and other diagnostics than serial correlation)


2 Answers 2


Since recently, you can use the DHARMa R package to transform the residuals of any GL(M)M into a standardized space. Once this is done, you can visually assess / test residual problems such as deviations from the distribution, residual dependency on a predictor, heteroskedasticity or autocorrelation in the normal way. See the package vignette for worked-through examples.

  • $\begingroup$ Thanks for pointing this out, will have a look at the package. $\endgroup$
    – Matthijs
    Jan 3, 2017 at 16:28

Under the assumption that you can order your residuals (typically in time or space), autocorrelation is just the correlation of the vector $[r_1, \ldots, r_{N-k}]$ with $[r_{1+k}, \ldots, r_{N}]$, for some lag, $k$. Normality doesn't much matter for correlation (cf., Pearson's or Spearman's correlation with non-normal data). It could matter for some tests of the significance of a correlation, but with autocorrelation you are typically more interested in the magnitude than its significance (by analogy, consider: Is normality testing 'essentially useless'?). If, for some reason you really wanted to test the correlation estimate, and didn't want to assume normality, you could bootstrap.


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.