2
$\begingroup$

I'm a beginner in time series analysis so I'm uncertain whether or not this data is stationary according to the ACF plot; I can observe a fast decay but on the other hand, a lot of the lines exceed -0.4 or 0.4. How would you conclude this?enter image description here

$\endgroup$
1
  • $\begingroup$ Clearly not stationary, with a correlation between consecutive points of about 0.7. $\endgroup$
    – Paul
    Oct 13 at 12:59
0
$\begingroup$

When using the ACF plot to diagnose stationarity, usually the rate of decay (into insignificant territory) in ACF values is of greater concern that the actual ACF values. If the ACFs decay slowly, that is usually a sign of non-stationarity.

Without context of your data, and due to your ACFs' wave-like behavior breaching the significance range and high lag 1 ACF, there is evidence of non-stationarity; but I'd advise using a statistical test to be sure.

There are several statistical tests for stationarity, including:

  • The Ljung-Box test for non-zero correlations at lags 1-20. The null hypothesis is that the series is non-stationary. Small p-values suggest that the series is stationary.

  • The Augmented Dickey–Fuller (ADF) t-statistic test: The null hypothesis is that the series is non-stationary. Small p-values suggest that the series is stationary.

Personally, I use the ADF test most frequently.

$\endgroup$

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.