2
$\begingroup$

I've created a DLM (using the dlm package in R) with a linear trend (dlmModPoly(2)) and a seasonal component (dlmModSeas(seasons)). After filtering, the residuals look Gaussian according to a qqplot. The confidence band around the estimates looks very good as well.

However, the acf plot of the residuals shows definite autocorrelation. I assume I need to improve my variance estimates from dlmMLE? Does anyone have any suggestions on where to start?

$\endgroup$
  • $\begingroup$ why don't you share your data , your mode, tour residuals and your analytics and perhaps we can help you. $\endgroup$ – IrishStat Dec 14 '11 at 21:44
1
$\begingroup$

Look at the ACF of the residuals, and decide what is missing in your model. Assuming the trend and seasonal part are already well accounted for with dlmModPoly(2) and dlmModSeas(seasons), you may want to to enlarge the model with an ARMA part. Look at the function dlmModARMA. Aside from the documentation of the package, you may want to look at the book by Petris et al. Dynamic Linear Models with R, p. 118.

$\endgroup$

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.