Conditional Forecasting Task

Jump to: navigation, search

A Conditional Forecasting Task is a multi-predictor forecasting task with a sequential mixed multi-predictor dataset.



  • (Kunst, 2012) ⇒ Robert M. Kunst. (2012). “Econometric Forecasting - Conditional forecasting." Lecture Notes.
    • Economists are often interested in forecasts for [math]x_{t+h}[/math] that assume [math]x_s, s \le t[/math] as known as well as the values of other variables [math]y_s, s \le t + h[/math]. The solution appears to be [math]E\left(x_{t+h} | x^t_{-\infty} \cup y^{t+h}_{-\infty}\right)[/math].
      • The assumed values [math]y_{t+1},..., y_{t+h}[/math] may be incorrect;
      • Any dynamic model that views [math]x_t[/math] as a function of [math]x^{t-1}_{-\infty}[/math] and [math]y^{t}_{-\infty}[/math] may miss the reaction of y to past x (feedback problem, open loop, weak and strong exogeneity);
      • Changing the generation mechanism for [math]y[/math] relative to the observations may affect the reaction of [math]y[/math] (super exogeneity).