Variation and prediction

In earlier pages of this section, we treated unexplained variation in data as 'noise' — a nuisance that cannot be avoided and that only serves to complicate the analysis of data.

This is not totally true. Variation can be interesting in its own right, especially when we are interested in predicting the future. If we can assume that there will be no systematic change to the process that generated the observed values in our data,

... the proportion of times that an event happened in the past indicates the chance that it will happen in the future.