A Decision-Theoretic Approach on Deciding when a Sophisticated Forecasting Technique is Needed

This note describes a decision-theoretic approach for deciding on when it is worthwhile to obtain a sophisticated statistical forecasting model and when an unsophisticated approach suffices. By developing a forecasting loss function for an example, the prior cost of uncertainty is first determined. Based on this cost, it is decided whether the sophisticated model should be obtained. If this is decided in the affirmative, the posterior cost of uncertainty is determined and the net gain from this decision is obtained.