[gus_massa]

The median is more robust than the average, and the average that is more robust than the maximal or minimal.

* One strange case/rat can modify the maximal or minimal a lot (For example, some kill a rat with a gun or you are "lucky" and get a Methuselah rat.)

* A strange event also can modify the average a lot (If someone kill a rat immediately at the beginning of the experiment, the "new" average is approximately (n-1)/n times the "original" average.

* But some unusual problem doesn't modify the median too much. The median changes from the value of a rat to the value of the next/former rat, which is usually very close. But it doesn't matter how small or big the strange case is, the median is essentially fixed between two values, so it is difficult to modify too much.

The real problem is that 6 is a small number and it is difficult to get good statistical result with only 6 cases.

[sanxiyn]

The study is not using median, it is using EML(estimated median lifespan), which is calculated by Kaplan-Meier estimator. This actually seems to be standard in lifespan studies.

http://en.wikipedia.org/wiki/Kaplan-Meier_estimator