I am analyzing the wind speed data which is a time series. The ARMA model works well on the data. But the same model fails to give good results when I difference the series. The Ljung-Box test gives p-value less than 0.05 (very close to 0). Does this mean that wind speed data may be considered a stationary time series?

Edit to the original post. The data was collected at 10 minute intervals. The plot is given below: enter image description here

  • $\begingroup$ I don't think that can be answered in the abstract. It might well depend on geographical location, for instance. Seasonality on year and 24h cycles is a possibility. Can you show some plots? $\endgroup$ – kjetil b halvorsen May 19 '17 at 17:40
  • $\begingroup$ Have a look at researchgate.net/publication/… The plot is difficult to analyze only visually, in addition can you plot an autocorrelation function? It could be long time dependence, making it very difficult to assess stationarity from a short series you shown. Calculating Hurst exponents could be useful. $\endgroup$ – kjetil b halvorsen May 23 '17 at 10:34
  • 1
    $\begingroup$ Please merge your two accounts (account1 and account2) -- you can't edit from a different account than the one you posted from. $\endgroup$ – Glen_b May 23 '17 at 11:41

This cannot be answered abstractly, we will need much more information. Relevant information needed to answer the question includes:

  • Frequency of observations. There might be yearly seasonality or 24-hour (daily, diurnal) cycles. If your data is means over 24-hour periods the last need not be considered.

  • Geographical location. Most places there would be a strong yearly cycle, but there might be exceptions to that.

Could you please augment your post with some plots, and the other auxiliary information?


You should try with a stationarity test "Dickey-Fuller",

In statistics, the Dickey–Fuller test tests the null hypothesis of whether a unit root is present in an autoregressive model. The alternative hypothesis is different depending on which version of the test is used, but is usually stationarity or trend-stationarity.

There are several tools in python and R to solve this problem..

Also the "Augmented Dickey–Fuller test" is a more robust test which is used to the same purpose.


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.