rationality and randomness http://awmitchell.posterous.com a personal blog of eclectic interests posterous.com Mon, 21 Sep 2009 13:53:00 -0700 Scanning Dead Salmon in fMRI Machine Highlights Risk of Red Herrings http://awmitchell.posterous.com/scanning-dead-salmon-in-fmri-machine-highligh-0 http://awmitchell.posterous.com/scanning-dead-salmon-in-fmri-machine-highligh-0

Some researchers have created some fMRI images which, at first glance, show that a dead salmon had a brain response when asked to interpret images of human emotion. Funny? Hell yeah! But instructive also.

Bennett’s point is that a suite of methods known as multiple comparisons correction can allow researchers to maintain most of their statistical power while keeping the danger of false positives at bay.

The work highlights that brain science is highly data-driven and statistical now. Although the visualizations — usually some orangey spots on an otherwise dark brain scan — seem simple, the data collection and interpretation that go into producing them is intense.

Vul, who published a controversial paper earlier this year that was critical of some statistical methods used in the field, said he appreciated that Bennett was also trying to do some “internal policing” to make fMRI practitioners’ methods as rigorous as possible.

Unfortunately...

Bennett’s paper has been turned down by several publications, but a poster on the work received an appreciative audience at the Human Brain Mapping conference earlier this summer. Neuroscience researchers have been forwarding it to each other for weeks.

via wired.com and thanks to bfchirpy's blog for the link to this item.

It's distressing that publications continue to have a bias towards papers with positive results. Negative results, including false positives, can be even more instructive. On the flip side, the informal networks are spreading the story anyway.

Permalink | Leave a comment  »

]]>
http://posterous.com/images/profile/missing-user-75.png http://posterous.com/users/4aQUIzR5baEN Andrew Mitchell awmitchell Andrew Mitchell