2
$\begingroup$

how can one handle a time series with the Chow's test (in order to find a structural break) so that the assumption of independent model errors holds? I'm using the R function chow.test {gap}

$\endgroup$
  • $\begingroup$ I am also interested in this question, as the autocorrelation will cause the Chow test to be more likely to detect structural breaks that don't actually exist, so a way to compensate for this would be very useful. $\endgroup$ – Dikran Marsupial Dec 19 '13 at 13:21
  • 1
    $\begingroup$ I'v found an interesting article on this topic. The title is "The Chow test with serially correlated errors" by Isabella Consigliere. This article is archived in Jstor. For your reference, the detailed information of this article's publication is the following: Anno 89, No. 2 (aprile-giugno 1981), pp. 125-137 Published by: Vita e Pensiero – Pubblicazioni dell’Università Cattolica del Sacro Cuore Article Stable URL: jstor.org/stable/41624786 $\endgroup$ – midtownguru Jan 22 '15 at 4:59
2
$\begingroup$

I implemented the CHOW Test in AUTOBOX by forming an appropriate ARIMA structure and then simply using all estimated parameters (both regressor coefficients and ARIMA coefficients) in the test. In fact in the case of just 1 group/class this test is used to find the point in time where the greatest contrasts exists thus testing the hypothesis of constant parameters over time. You can do the same in R or any other programming language just be careful about outlier effects and possible non-constant error variance. See How can I programmatically detect segments of a data series to fit with different curves? for an application of the CHOW test to detect possible regime changes.

| cite | improve this answer | |
$\endgroup$
  • 1
    $\begingroup$ Congratulations on reaching 10,000 points, I have personally benefited on reading your posts on time series forecasting and outlier correction methods. $\endgroup$ – forecaster Jun 14 '15 at 1:01
  • $\begingroup$ I really appreciate your frankness and your very evident scholarship. $\endgroup$ – IrishStat Jun 14 '15 at 10:23

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.