I'm creating a dynamic regression model in which macroeconomic indicators are predictors/features in the model. I need to forecast these features n-steps into the future. I am planning to decompose the time series using statsmodels package from Python. Next, i will take the trend and seasonal components and extrapolate the series n-steps ahead and then combined them for use in the test dataset. Is this the correct approach to extrapolating predictors, absent any other forecast knowledge about their future value or are there better approaches. Despite viewing many papers and articles on forecasting methodology, there is a strange dearth of material on forecasting the predictors. thanks

  • $\begingroup$ it's one possible approach when you have a stable trend and seasonality $\endgroup$ – Aksakal Jun 17 '19 at 17:57
  • $\begingroup$ stats.stackexchange.com/questions/410082/… presents an example of user predictors can be used . If the predictors have to be predicted then the uncertainty in those predictions need to be incorporated into the final forecast of Y possibly via Monte-Carlo methods. $\endgroup$ – IrishStat Jun 17 '19 at 18:14
  • $\begingroup$ the method for predicting the predictors should be based upon the information content in the history of each predictor. $\endgroup$ – IrishStat Jun 17 '19 at 18:40
  • $\begingroup$ Thanks, Aksakal $\endgroup$ – WON_Eric Jun 18 '19 at 15:07
  • $\begingroup$ Thank's IrishStat $\endgroup$ – WON_Eric Jun 18 '19 at 15:07

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Browse other questions tagged or ask your own question.