Comparing Probability Forecasters: Basic Binary Concepts and Multivariate Extensions

摘要: Abstract : In the applied forecasting literature much attention has been lavished on questions about evaluation of probability forecasts, and subjectivist view invoked to aggregate forecasts over a diverse set events or statements. One criterion in such evaluations is that calibration: statements considered we ask if x percent those assigned being correct prove be correct, for each value x. From this perspective, weather forecasters generally have found perform well. What especially helpful they make long sequence (e.g. rain given day), thus it makes sense think functions associated with forecasts. paper focus comparing forecasters, refinement, which goes beyond calibration.