The scatterplot is one of the simplest but most useful ways of evaluating forecasts of a continuous variable. All the data are shown in a convenient graphical format, normally with the forecast on the abscissa and the observation on the ordinate. It is most convenient if the same scale is used for both forecast and observations. Then perfect forecasts for which forecast=observed are represented by a diagonal at 45 degrees. The forecast=observed line is usually included on the graph to facilitate interpretation. The individual cases of the verification sample are most often plotted as dots or other symbols. If the data sample is very large, sometimes subsets of the data are represented as circles centered at the mean forecast and observed value, with a radius proportional to the number of cases represented in the subset.

The five questions following the sample scatterplot below will lead you through the interpretation of scatterplots.