What Dead Salmon Tell Us About Statistics

Greetings Students of Science!

In class, we discussed the importance of using the appropriate statistical analyses and controls, but it may surprise you to find out that using a dead salmon in a research study can actually lead to a scientific revolution!

In 2010, a study published in the 'Journal of Serendipitous and Unexpected Results' (yes, that is a real scientific journal) showed how important it is to properly analyze your data. In their study, Bennett et al. put a dead salmon in an fMRI machine, showed the salmon a serious of visual images, and then analyzed the fMRI results in several different ways. Typically, fMRI data is analyzed by comparing changes in blood flow
between up to 130,000 different subareas of the brain (called voxels). However, all the different comparisons made exponentially increases the chance of finding a false positive result (e.g. a significant result when there isn't one). While most fMRI researchers correct for these multiple comparisons, at the time that Bennett et al. did their study, some fMRI researchers did not.

Dead Salmon 'brain activation.' Image courtesy of
http://blogs.scientificamerican.com/scicurious-brain/2012/09/25/
ignobel-prize-in-neuroscience-the-dead-salmon-study/

To address this issue, Bennett et al., analyzed their fMRI-dead-salmon data in two different ways (either controlling for multiple comparisons, or not). The authors found that when their data were analyzed without correcting to the use of multiple comparisons, the dead salmon's brain showed significant activation in response to various visual stimuli. Surprisingly, the reaction from the scientific community was mostly positive, and encouraged many researchers to rethink the way they analyzed their data.

If you want to read more about this 'fishy' tale, please click here.