This is my very first time building time series forecasting and i'm currently trying ARMA in python. I'm stuck in building my ARMA (ARIMA(p,0,q) model because of there's no significance at all in my ACF and PACF plot. I have read several articles about ARIMA but all of them at least shows significant correlation in their ACF and PACF plot. So for my case, i don't know what to do since this is my first time building times series forecasting model. My data is very stationary so i thought i could go on to plot the ACF and PACF and to build the model. But now i start to doubt if ARMA suits my problem.

  1. What should i do if i could still go on building the ARMA model?
  2. If ARMA is not for my data, what other algorithm should i use?
ADF Statistic: -7.654896
p-value: 0.000000
Critical Values:
        1%: -3.508
        5%: -2.895
        10%: -2.585

My PACF ACF plot


Seasonal Decompose


Fit the ARMA to my data, red line is the predicted, blue is the observed


  • $\begingroup$ Pulses , level shifts, seasonal pulses and local time trends can mask the identification of arima models . Why don't you post your actual data and I will try and help further . $\endgroup$
    – IrishStat
    Commented Apr 28, 2020 at 22:58

1 Answer 1


I am not sure what you mean by significance. It looks like you have no AR or MA that matter which is what you want to have. You should run a box ljung test and see if you have any serial factors that still need to be addressed. ARIMA assumes that variation does not change over time and that may not be true in your model. It may take a different type of model like GARCH, but these are more complex than ARIMA. Which by the way is not simple, reading a few articles on it may not be enough. Have you tried AUTOARIMA in R?

  • $\begingroup$ If all you want is to forecast you might try exponential smoothing. According to the literature, and my experience, they have good track records of being right and are far simpler to do than ARIMA which is much an art form as a science as my first instructor pointed out. It is not an easy place to start especially if forecasting is all you care about. $\endgroup$
    – user54285
    Commented Apr 29, 2020 at 23:29

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.