Has anybody played around much with Google’s Correlate tool? Quite amazing, in a frightening sort of way. I found it surfing from this similarly amusing, but less thorough, post. I can come up with no adequate theory to explain the nearly .73 correlation between my scribbled line and searches for “home videos clips” on Google.
However, using a steadily decreasing line, the correlation is mostly with searches for now-obsolete web technologies (how many people now search for “web page” on google?), which indicates, I think, the fact that the meaning of google itself has changed over time.
One Comment
I do use a couple of their examples (“mittens”) to make exactly the point your title suggests: to teach students about correlation and causation.
I drew a distribution I know without thinking about the meaning of the axes and that yielded nonsense. There are not a lot of problems I care about that have “a count of web searches” and “over the last decade” as relevant characteristics. I suppose that’s what you’re getting at by pointing to the importance of an interest in “the meaning of google.”