n
nicefa.
Advanced Proof

Derivation of the Formula for a 95% Prediction Interval for an AR(1) Forecast

A common pitfall is confusing a prediction interval (for a single future observation Yt+h Y_{t+h} ) with a confidence interval (for the conditional mean E[Yt+hFt] E[Y_{t+h}|\mathcal{F}_t] or a model parameter). Students also frequently omit or miscalculate the cumulative variance contribution from past innovations, especially for h>1 h > 1 .
Institutional Reference: Time Series Analysis
View Full Proof →