- Joined
- May 16, 2006
- Messages
- 2,724
I believe he went with median to reduce the effect that any fluctuations to an extreme would have on the testing. Using your numbers, theonew, in each of 2 tests, (1) 40g and (1) 60g cut is an extreme measurement, a fluke, statistically speaking. Using the median reduces the weight that the fluke(s) carries on the measurement as a whole. I am just guessing, but you have to remember that Vassili is a statistician, so tries to reduce the variables with numbers.
My example was weighted to the extreme, so change the number of cuts on the upper and lower ends a bit and you are still getting an unnecessary degree of inaccuracy I would think. A fluke in my mind would be if you had one or two cuts which were either 20 grams or 80 grams, using the example numbers I gave and this may be what Vassili is seeing. But from the few examples he's given it looks like the range is constricted to a 20 gram variance.