A Decomposition of the Correlation Coefficient and its Use in Analyzing Forecasting Skill

Abstract

Estimates of several components of forecasting skill can be obtained by combining a skill-score decomposition developed by Allan Murphy with techniques for decomposing correlation coefficients that have been employed in research on human judgment. The decomposition of the correlation coefficient requires knowledge of the information or ?cues? used by the forecaster. When the cues are known, it is possible to estimate the effects of uncertainty and the forecaster's consistency and use of the cues.