From: Arthur Lewbel <[email protected]>
Date: September 19, 2009 9:57:57 AM EDT
To: [email protected]
Subject: false positives
Faculty,
I just came across a wonderful illustration of the problem that if
you estimate too many parameters, some will be significant just by
random chance. This is the link to a poster that was presented at
the 15th annual conference of the organization of human brain
mapping: http://prefrontal.org/files/posters/Bennett-Salmon-2009.pdf
In this study, the authors put a Salmon in an MRI machine. They
showed the fish pictures of human faces, and recorded the responses
in thousands of tiny areas within its brain. A small number of
statistically significance responses (p value less than .001) were
found. What's most amazing is that the fish was dead at the time.
Arthur
=================================================+
Arthur Lewbel [email protected] http://www2.bc.edu/~lewbel/
phone: (617)-552-3678 fax: (617)-552-2308
Dept. of Economics, Boston College, Chestnut Hill, MA 02467, USA
==================================================
