Learning paths

Trend, Range and Scale

Three papers that turn out to be one argument. The first counts how long trends actually last and finds a distribution too skewed for an average to describe. The second counts how often range edges hold and finds the same distribution mirrored, with the fat tail on the other side. The third shows that both counts were written in bars, which makes the bar width part of the definition of every number they produced rather than a lens on it.