I am completely new to R, time series and ARIMA, so bear with me. I have a time series data from 2018 to 2020 that shows an intervention in early 2020. When plotted, the graph quite dramatic plummets post-intervention. I have done an ARIMA forecast (to the best of my ability) for the pre-intervention period and when comparing the ARIMA forecast with the post-intervention period (actual data), there is a massive difference. My questions are:

  1. What do I actually do with the ARIMA now? My data is stationary (0,1,1) and an Ljung box test shows a p-value of 0.2881, which I'm assuming is good. But how do I actually show that the intervention has had indeed an effect?
  2. Is there a way to plot the forecast together with the actual data? I've tried different ways and so far have had no luck.

Thank you


1 Answer 1


I suggest using auto.arima if you are new to ARIMA. I use it and I have studied ARIMA for a while. :) It is FORECAST package in R. It quickly chose the model, something the classical methods struggle with when you have mixed models (that is both AR and MA).I am not sure what you mean by a massive difference. A massive difference in what? It could be there was a change in the process after the intervention. There are several options including interrupted time series, segmented regression, and test to see if there was a change through a chow test. https://en.wikipedia.org/wiki/Chow_test I should warn you I am a data analyst not a statistician so be cautious about my advice. :) It is what I have seen others recommend.


Your Answer

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

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