METplus Practical Session Guide (Version 5.0) | Basic Verification Statistics Review > Multicategorical Forecasts

Multicategorical Forecasts

In practical applications of forecast verification, it’s often of interest to look at more than two categories and cannot be reduced to a binary, “did the event happen or not, and was the event forecasted or not”. Restricting forecast and observation datasets to binary options leads to the loss of important information about how the forecast performed (e.g., when multiple categories are combined into just two categories). For example, if the forecast called for rain but snow was observed instead, it can be useful to analyze how “good” the forecast was at delineating between rain, snow, and any other precipitation type. Luckily, the transformation of statistics from supporting binary categorical forecasts to the second group of forecasts, multi-category, is fairly straightforward.