News for hackers. Stuff that matters.
Re:On Haskell, Ruby, and Cards Against Humanity by zxdfghj on Monday April 27, 2015 @11:56AM (#9447096)
You can meaningfully count up the total number of fingers that five people have between them, and divide it by five.

You cannot meaningfully add together how funny five people are and divide it by five. However, you can put five people in order of how funny they are, and point to the middle one.

The only appropriate concept of average for ordinal data (e.g. how funny someone is) is the median, from which it follows by definition that 50% of people are less funny than average.

PS - if it is true that it "cracks you up" every time you hear "the 50% of people are above/below average fallacy", I suggest that you might be one of the 50% of people who have a below-average sense of humour ;)

Re:On Haskell, Ruby, and Cards Against Humanity by taeric on Monday April 27, 2015 @12:16PM (#9447219)
I would wager that among any group of five people that they all agree that one of them is the funny one.

I would extrapolate further and say since there is a tiny number of comedians performing any night that there is both a non-even distribution of humor and that it is weighted towards fewer people having it. Consider, if it were so common, would the rest of us be as willing to pay to hear someone else perform?

And to be fair, I don't think that would be evidence in itself. I would be curious to see actual studies into this. At least a little.

Re:On Haskell, Ruby, and Cards Against Humanity by dragonwriter on Monday April 27, 2015 @12:33PM (#9447352)
> The only appropriate concept of average for ordinal data (e.g. how funny someone is) is the median,

Incorrect. For the four usual kinds of measures, the following types of "averages" are generally useful:

Nominal: Mode

Ordinal: Mode, Median

Interval: Mode, Median, Arithmetic Mean

Ratio: Mode, Median, All Means (arithmetic, geometric, harmonic)

> from which it follows by definition that 50% of people are less funny than average.

Nope, that only follows by definition for ordinal data with the additional assumption that no items have equal rank.