In a behavioral study, a neurosurgeon, a pathologist and an epidemiologist are each observed while they study a can of sardines.
The neurosurgeon enters the room, and emerges 5 minutes later. At first glance, everything is as it was previously. Closer examination reveals that the can has been opened at the top rim very precisely, and the top replaced. On the counter is a piece of paper with “can of sardines” scribbled on it.
The pathologist enters the room (there is a fresh can of sardines) and does not emerge until an hour later. The can of sardines has been smashed open, and bits of sardine, metal and oil are strewn across the table top. There is a two page report on the table, that identifies the type of oil, type of metal, quantity and weight of each. The sardines have been counted, and the report confirms that each of them is dead.
The epidemiologist stays in the room for 8 hours and the investigators are growing anxious. Finally, he comes out – the can of sardines is untouched, but there is a 20 page report on the table. This report contains a preamble that discusses the potential provenance of the can, and a main section that details various statistical methods for estimating the number of sardines in the can, and provides several actual estimates (all different), and confidence intervals that go with those estimates.
Source: An Epi 1 lecture at Columbia University in Sept. 2001 (embellishment added), available on YouTube here: http://www.youtube.com/watch?v=jCK2mflwESM