Friday, June 4, 2010

What Stats Are Good For?

Identifying trends!  While correlation does not equal causation, a temptation to which many folks using quantitative methods succumb, numbers are pretty good for identifying trends if used correctly.  This piece in Slate [HT to Pete] talks about three non-trends, showing that the echo chamber that is today's internet/media interface tend to repeat the same story and tend to have no sense of the past. 

Some things are new and increasing in frequency, some things are old and occurring at a fairly consistent rate, and somethings are declining.  How do we know?  Well, watching TV news is probably not the way to figure it out as they have incentives to play up anything as new. 

The problem again is one of cognition, as we tend to believe what we expect to be the reality.  My daughter this morning was miffed about a letter to the local paper comparing the Liberal members of the Quebec assembly to Ostriches for veritably sticking their heads in the ground while their party proposes legislation that harms the interests of the Anglophone voters.  Why was she miffed?  Because Ostriches do not stick their heads in the ground. 

I guess what she recommends is that more folks watch Mythbusters.

No comments: