I'm particularly interested in plotting residuals against fitted values, and residuals against predictors. Often times I need to make boxplots of the residuals conditional on predictors.

I'd be interested in functions that include other/multiple diagnostics as well, e.g. ceres plots, qqplots.

The car package has some nice diagnostic functions but they are only for objects of class lm and I often use lmer.

  • $\begingroup$ Did you check Gelman's ARM package? I seem to remember there were pretty nice functions in there. (Would have to check again, though.) $\endgroup$ – chl Nov 1 '11 at 8:59
  • $\begingroup$ I certainly could be wrong, but I wouldn't be surprised if you need to roll your own because exactly how to treat the different levels of residuals you get from a mixed model may depend on how your model is set up. I remember Gelman having functions to simulate data from the model to see if it looks like the original. Very good for diagnostics, though not quite what the OP has in mind. $\endgroup$ – Aaron left Stack Overflow Nov 1 '11 at 16:37

Not really about partial residuals plots, but I came across the influence.ME package which seems to address the last point, if we consider diagnostic measures to extend to influential observations.

I also found some illustrations that were apparently done in R in the following paper:

Nobre, JS and da Motta Singer, J (2007). Residual Analysis for Linear Mixed Models. Biometrical Journal, 49(6), 863–875.

(But check out the following slides, Residual Analysis for Linear Mixed Models, by one of the authors.)

| cite | improve this answer | |

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

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