This poster, by Bennett, Baird, Miller, and Wolford, provides a memorable reminder that you have to do a statistical correction for multiple comparisons when you datamine a large number of things for correlation.
“The task administered to the salmon involved completing an open-ended mentalizing task. The salmon was shown a series of photographs depicting human individuals in social situations with a specified emotional valence. The salmon was asked to determine what emotion the individual in the photo must have been experiencing.”
Voxels in the data were searched to find voxels which show a statistically significant correlation with the experimental condition. 16 such voxels were found (without doing any correction). As the poster says, “Across the 130,000 voxels in a typical fMRI volume the probability of a false positive is almost certain”. The poster recommends correcting with either “overall false discovery rate (FDR) … based on a method defined by Benjamini and Hochberg (1995)” or “overall familywise error rate (FWER) …. using algorithms originally devised by Friston et al. (1994)” (with either correction, no voxels were found).