Skip to content
Time-Series
EconometricsIan Helfrich PhD ↗

Lecture 15 summary · April 30, 2027

Defend a forecast in an unfamiliar setting

Complete chapter · Laboratory ch15 · Practice · All summaries

The economic question

Transfer the argument, not its assumptions automatically. An unfamiliar regional financial setting still requires a target, information clock, dynamics, uncertainty, and defensible action.

Central relation

Y^t+h∣t=g(It;θ^t),\widehat Y_{t+h|t}=g(\mathcal I_t;\widehat\theta_t),
E[(Yt+h−Y^t+h∣t)2∣It]\mathbb E[(Y_{t+h}-\widehat Y_{t+h|t})^2\mid\mathcal I_t]

Individual problem

Future records change while admitted observations remain fixed, and the forecast changes. What has been established? If monetary units multiply by one hundred, what happens to errors and variances?

Worked reasoning

Future dependence or hidden global state violates the declared boundary. Errors multiply by one hundred; squared errors and variances by ten thousand. Dimensionless persistence is unchanged.

A consequential failure

Complexity is defended instead of information and later evidence, or regional predictive dependence is interpreted as a causal policy effect.

Repair

Run future, unit, rank, covariance, and domain checks. Separate prediction from causality and keep the benchmark when its deployment evidence is stronger.