Allen Downey, Probably Overthinking It blog A related statistical trap not covered in this chapter: why samples you draw are often biased toward overrepresenting high-frequency or high-visibility events. Relevant for any analyst trying to understand customer behavior from observed data.