Announcement

Collapse
No announcement yet.
X
  • Filter
  • Time
  • Show
Clear All
new posts

  • Recursive ARDL Forecasting

    Good evening all,

    I am using Stata 18.

    I am trying to run an ARDL model that predicts one step ahead, replaces the Y-variable with the one-step ahead forecast, and reruns the ARDL model on the next step ahead forecast to then predict a further one step ahead forecast (and so on, and so on).

    I am not able to share a snip of the actual data (confidentiality), but am able to provide a very simple analogue that I hope will be sufficient:

    Code:
    * Example generated by -dataex-. For more info, type help dataex
    clear
    input float time byte(y x1 x2)
    2020 28 63 15
    2021 79 89 57
    2022 84 28 84
    2023  . 61 42
    2024  . 10 41
    2025  . 67 94
    end
    The ARDL model would be of the form:

    ARDL y x1 x2, trend(time)

    Using predict, I could safely predict 1 period forward (i.e., to 2023 in this case, given that I have forecasts for x1 and x2). However, I would like the 2023 number to now reflect as a new Y entry, before running the same regression again (but now with the "updated" Y entry up to 2023, and the subsequent prediction predicting for 2024).

    Many thanks and warmest regards,
Working...
X