A Statistical Methodology for Quantifying Improvement in Predictive Capability: Key Features and Benefits
In References [1, 2, 3], we proposed a statistical methodology for quanti fying improvement in predictive capability. In this note, we summarize the key features and benefits of our approach. We begin with a brief summary of our methodology. Broadly speaking, our approach models the predictive process statistically at a high level. Ac cordingly, we assume the existence of a body of experiments, the results of which have been (or will be) predicted. We can then treat the discrepancy between a prediction and the corresponding experimental result as a datum and the collection of such data informs us about the statistics of the prediction process.