I have a nonlinear least squares problem, but the errors are correlated. I could use R's nls function to do the regression if the errors were independent, but I don't know the right way to handle correlated errors. Is there a standard procedure for this? Is it implemented somewhere?

  • 1
    $\begingroup$ Do you know anything about the nature of the dependence? For example, is the correlation between two errors the same for every pair of errors? Or could it depend in some way on the predictors? Also, can you tell us anything about the nature of the nonlinearity? $\endgroup$ – Michael Hardy Aug 1 '12 at 19:09
  • $\begingroup$ @MichaelHardy I believe the errors are autocorrelated. $\endgroup$ – BenH Aug 1 '12 at 19:46
  • $\begingroup$ You should post this on the stats SE, I bet there's someone that could help there. $\endgroup$ – TheHopefulActuary Dec 3 '12 at 7:03
  • $\begingroup$ @MichaelHardy You can try GARCH or EGARCH $\endgroup$ – AnilB Mar 14 '13 at 16:01

Your Answer

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

Browse other questions tagged or ask your own question.