Open AccessThree new diagnostic verification diagrams
Author(s) -
Wilks D. S.
Publication year - 2016
Publication title -
meteorological applications
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 0.672
H-Index - 59
eISSN - 1469-8080
pISSN - 1350-4827
DOI - 10.1002/met.1561
Subject(s) - computer science , probabilistic logic , diagram , representation (politics) , graphical model , calibration , factorization , joint probability distribution , data mining , conditional probability , algorithm , artificial intelligence , statistics , mathematics , database , politics , political science , law
Forecast evaluation is inherently multidimensional, so that graphical representations of verification datasets, which can simultaneously show multiple dimensions of forecast verification, are helpful for diagnosing the strengths and weakness of the forecasts. This paper presents three new diagnostic verification diagrams for settings that previously lacked them. The H‐F diagram is a graphical representation of the likelihood–base rate factorization, and the T‐M diagram is a graphical representation of the calibration–refinement factorization, of the joint distribution of forecasts and observations in the setting of non‐probabilistic forecasts for dichotomous events. The Conditional Boxplots diagram represents graphically the calibration–refinement factorization for probability forecasts of continuous outcomes, where each forecast is itself a continuous probability distribution function. The three new diagrams are described, and illustrated with example verification datasets.