Statistics is Applied Science Fiction

I’m enjoying the discussion coming out of Alberto Cairo‘s online data visualization course.

Bryn Williams, in a comment on thinkers & creators who read comics & sci-fi for inspiration:

“…a familiarity with imagined alternative worlds makes philosophy an easier path to tread when posing counterfactuals and thought experiments…”

My response:

And not just philosophy or data visualization — I think statistics could be presented as a kind of “applied science fiction.” When you perform a hypothesis test of whether some parameter is 0, you

1. assume it *is* 0,
2. imagine what kinds of data you would probably have seen under that assumption, and then
3. if the real data you *did* see is unlikely under that assumption, decide that the assumption is probably wrong.

It’s just like in SF where

1. you imagine a possible alternate reality (say, Joe discovers a talent for dowsing),
2. you explore the consequences if that possibility were true (Joe becomes rich from oil prospecting), and
3. in the best cases, readers can draw lessons about our actual reality from this thought experiment (http://xkcd.com/808/).

(XKCD is, of course, a great comic for both SF and datavis. See also this recent SMBC for another amusing exploration of “If this claim were true…”)