[bambax]

It's not just Mary Jones. The controversial (to say the least) French medical professor Didier Raoult last year showed he didn't quite grasp the difference between median and mean. There was some mockery on Twitter but no journalist ever called him up on it.

[brutusborn]

I couldn't believe an Academic could be this ignorant, so I had to look it up. I found it but only basic french. I can't find a translation, does anyone know of one? https://mobile.twitter.com/cubic_logic/status/15410902478613...

[d1sxeyes]

EN: '[...] some things to put together. You see that if we look at the mean or the median, the mean is when a distribution is like this [indicates a bell curve with his hands], we can calculate a mean, but when it's like this [wiggles his finger], you can't calculate a mean, you do the, the, the bar which separates them in half. And so [...]'

FR: '[...] d'objets de mettre ensemble. Vous voyez que si on regarde la moyenne ou la médiane, la moyenne c'est quand une distribution est comme ça, on peut faire une moyenne, quand elle est comme ça on n'a pas [...] pour faire une moyenne, on fait, le, le, le, la barre qui sépare à moitié les uns les autres. Et donc [...]'

There's a short phrase I can't quite catch at about 9s... which literally translates as 'we don't have [...] to do a mean', so I'm inferring that he means 'you can't calculate a mean' from context (previous clause 'on peut faire une moyenne' => 'we can do a mean' => 'we can calculate a mean').

[alain94040]

Actually, I believe he is probably correct in that case. He is discussing whether the mean is meaningful depending on the shape of the distribution. It's probably not fair to summarize his explanation as saying he doesn't understand mean and average.

HN's guidelines recommend to assume the best.

[d1sxeyes]

I would agree that you should always try to assume the best (and I didn't actually suggest he doesn't understand), but he certainly explains it clumsily, and his explanation does not illustrate the difference between a median and a mean, or when one would be more meaningful than the other.

[pkd]

I don't think it's the best explanation, but I do think he understands that mean as a measure of central tendency is not that useful when distributions are skewed or multimodal (the squiggly line gesture).

[d1sxeyes]

Yes but:

1. If data is multimodal (wiggly line), the median isn’t a great measure of central tendency either.

2. Perfectly normally distributed data will have the same mean and median.

I agree people have jumped on this as evidence the guy doesn’t understand when he probably does, but what he says is a long way from the truth.