This question already has an answer here:

I have time series data. Not sure whether the series is stationary or not. I applied adf.test and got a p-value much higher than 0.05 which I think means that the data is not stationary. I took differencing once to check the p-value again. And I see that p-value has gone below 0.05. Can I assume that the data is stationary and ready for forecast? Is there any other way to confirm this in 'R'?


marked as duplicate by Stephan Kolassa, Peter Flom Jul 14 '18 at 12:17

This question has been asked before and already has an answer. If those answers do not fully address your question, please ask a new question.

  • $\begingroup$ Did you made a plot before to jump into computations ? $\endgroup$ – AlainD Jul 10 '18 at 18:32
  • $\begingroup$ Yes.. plotted ACF and PACF as well, I could see autocorrelation in the data. Once I took differencing, I don't see autocorrelation and adf.test also shows p-value less than 0.05 $\endgroup$ – Vijaya Patil Jul 11 '18 at 17:11

Yes and no. Yes, ADF test tells you that after differentiation your data became stationary. No, your data is not ready for forecasting yet.

Run your time-series model using stationary data first and then check for issues such as heteroskedasticity or autocorrelation before forecasting.

You can also examine residuals graph and ACF/PACF.

Forecasts are not accurate if heteroskedasticity/autocorrelation is present in the errors.

  • $\begingroup$ The ADF test requires that there are no pulses, level shifts, seasonal pulses and local time trends in the series that need to be treated. $\endgroup$ – IrishStat Jul 11 '18 at 13:25
  • $\begingroup$ @Justina: Thanks for your help! I don't see autocorrelation in residuals. I will divide the data into train and test as suggested by you to be sure of forecasts. $\endgroup$ – Vijaya Patil Jul 11 '18 at 17:03

Not the answer you're looking for? Browse other questions tagged or ask your own question.