2
$\begingroup$

I am modeling a time series using a multiple (dynamic) linear regression model. I suspect that at some point, the model no longer accurately predicts the true series. Is there a way to find the point where the true series and the model diverge?

Ideally I'm looking for an already implemented R-package.

$\endgroup$

1 Answer 1

1
$\begingroup$

There is an R package called changepoint.forecast available on Github here. It implements online changepoint models which look at the forecast residuals and check for changes within them, mainly changes/drift in the expectation and variance. It can be used with any model that produces forecasts which are expected to capture the mean and second-order structure (variance/autocovariance) of the data, including black boxes.

The package will be released on CRAN when the paper is further along the publication pipeline (it was submitted earlier this year). There is a talk describing the functionality here.

$\endgroup$
1
  • $\begingroup$ Thank you! That is exactly what I was looking for. Absolutely incredible that you found that. $\endgroup$
    – steinchen
    Jun 9 at 11:10

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.