I'm performing a cointegration test + OLS regression on 2 assets' prices. If I check the correlogram of the model residuals (training set, but also happens with the test set residuals) it looks like there is a perfect 1-lag autocorrelation in the residuals: enter image description here

enter image description here

And this is what my residuals look like:

enter image description here

What can be the cause of this? I'm assuming something must be wrong, but no idea what.

  • 1
    $\begingroup$ Is the cointegration of order 1? $\endgroup$
    – jbowman
    Feb 15, 2023 at 18:45
  • $\begingroup$ I take the first difference (I(1)) of the log prices at time t always (et - mean = P1t * coeff - P2t), I never use any kind of lag $\endgroup$
    – Hiperfly
    Feb 16, 2023 at 1:56
  • $\begingroup$ So you have a regression of $\ln(Y_t)$ on $\ln(X_t)$ where $Y_t$ and $X_t$ are prices? And the residuals looks like in the plot? Then your time series might not be cointegrated. $\endgroup$ Feb 16, 2023 at 10:34


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Browse other questions tagged or ask your own question.