Truly easy & getting straight to the point! Now the school lectures get overwhelming :). Thanks for the effort you make! P.S. Info for those who made the same mistake as me while forecasting: if you think that only the tale-part will get a predicting values, this would be wrong. Arima is indeed being made on the basis of the first 12059 observations or so and then these same observations plus all the others get an Arima-prediction. In this video it can only be seen while looking on browse tables after the prediction command and the dummy gen were both executed, since there was no complete plot with prediction for the whole time series.
Great question! I'd recommend doing the estimation and forecast within a 'foreach' loop, with the endpoint of the sample increasing by one unit each iteration. I don't have a tutorial for this exactly (yet!), but a general loop example can be found here: ru-vid.com/video/%D0%B2%D0%B8%D0%B4%D0%B5%D0%BE-sT3gjnd9v-k.html Let me know how it goes!
@@mikejonaseconometrics1886 Great video, thanks. Like other viewers Savannah and Denny, I would like to be able to make a series of Arima forecasts for t+1. It's quite easy for regression models (save betas, merge to data, generate forecast with linear formula) but to get a rolling estimate from Arima seems to require a deep knowledge of Stata's command language. A video or any pointers would have an appreciative audience. Thanks.
Amazing lecture. Thank you very much. Unfortunately I can only like once. ;-) What should I do if I want to make a forecast out of sample 4 steps ahead??
Thanks! To create a 4-step out-of-sample forecast you first have to add observations to the end of the sample using "set obs `=_N+4'. Then you can estimate the model and forecast using "predict forecastname,dynamic(N)", where N is the last observation number in the original sample.