Hacker News new | ask | show | jobs
by bradleyjg 1594 days ago
An average from a bimodal distribution occludes more than it reveals. These days we all have a bit of a fetish for what appears to be data driven analysis even when it involves something like this where it makes things actively worse.
1 comments

The hiding of the data’s distribution always makes me sad.

It is not costly to provide quintile or decile data, yet no one ever does when that is where the meat and pickles are. I assume most of the time, averages (and especially unspecified averages where you do not know if it is mean or median), are used to invoke emotions and land more clicks.

Comparing averages is nearly useless without knowing the distributions.
You mean _totally_ useless.