Hybrid Weighting Method

---

The different results in the two tables for samples A and B suggest that there might be value in introducing a new weighting process, a "hybrid" approach, that uses the Fixed Weights method if the quadratic programming model of estimating "optimal" weights by OLS gives two or more nonzero weights but uses the Simple Average weights if it only identifies one weight. This would have given an overall sum of root mean squared errors of 3905, the best possible result. This value is close to that given by the use of the actual best forecasting method, which remember cannot be identified until after the event. It saves 4.2% on the use of the best individual method on the basis of the previous 12 month's performance, the Highest Weight method. Practically, this hybrid method would seem easy to apply in practice as it only requires updating of the weights once per year.

---