extreme values

This term is used in two senses in the book.
The first describes anomalous outcomes in your data. For example, if 10 coins are tossed there is about a 1 in 50 chance of getting only 1 head or 1 tail. Really extreme values in results are relatively unlikely. If they are unexpected they are called outliers, and may arise by chance or from faulty equipment; if they are valid data points but just happen to be extreme, they are part of the random nature of the phenomenon being studied.
The second usage refers to the various parameters in a distribution or model. Even if you expect real values to be more central, experimenting ... playing! ... with more extreme values can often give you a sense of the data. For example, experiments where you toss (virtual) coins with very high bias or coin-to-coin correlation.

Used on pages 30, 38, 41, 81, 121