I've been given daily data and I've trained a SARIMAX time series model in Python so that I can predict daily data if given daily input.

However, I need to forecast on a monthly or weekly level, meaning my input would be in monthly form and not daily input. How do I go about forecasting with my current (daily) model in a different time interval?

Do I need to recreate my model so that it's trained on weekly/monthly data?


1 Answer 1


Develop your daily model taking into account day-of-the-week, day-of-the-month, lead and lag effects around holidays, level shifts, monthly effects, time trends etc. .

Now forecast out 1 period and generate a family of possible values say 1000.. call that simulation1 allowing for possible pulses to occur. Now do that for period 2 while incorporating increased uncertainty ... then ... do the same for period 30.

Now accumulate all 1000*30 forecasts and then sort them from low to high. Find the 2.5% value and the 97.5% value and you will have 95% confidence limits.

See my response to ARIMA model, daily data, weekly external regressor where I discuss monte carlo simulations and the concept of combining them / aggregating them .

  • $\begingroup$ by 'periods' are you referring to days of the month? after i get my 95% confidence limits, how do i know what values to assign to each day of the month (given monthly input)? $\endgroup$
    – tiffany123
    Jul 11, 2019 at 17:02
  • $\begingroup$ "How do I go about forecasting with my current (daily) model in a different time interval?" since you have daily data you can develop a daily forecast and using monte-carlo simulation create a family of forecasts for each day in the future and the generate a rolling sim for each day in the future. $\endgroup$
    – IrishStat
    Jul 11, 2019 at 19:07
  • $\begingroup$ are you clear on what I am suggesting ? $\endgroup$
    – IrishStat
    Jul 12, 2019 at 19:01
  • $\begingroup$ If you are happy with my answer , please accept it and close the question $\endgroup$
    – IrishStat
    Aug 15, 2019 at 7:59

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.