I have a linear AR(2) regression model of an time series that looks like this:

$y_{t} = \beta_0 + \beta_1 y_{t-1} + \beta_2 y_{t-2} + \beta_3 x_t + \epsilon_t$

I am modelling the contribution of the time series $\{ y_{t-1}, y_{t-2}, x_t\}$ on tomorrow's $y$, $y_{t}$.

I want to test whether these contributions change over time. I do a rolling regression over a period of ~10 years giving me in total a set of about 2000 beta parameters $ \{\hat{\beta_1}, \hat{\beta_2}, \hat{\beta_3} \}$ over time. Indeed inspecting the change qualitatively reveals that there is a trend in the value the parameters over time.

My question: How can I test formally that there is indeed a structural change in the parameters over time?


Chow test is what you want. Also, Google "parameter constancy".


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.