Voodoo correlations

Very often science can be somewhat trivial: First, a correlation is found between two properties that nobody was expecting. Second, someone proposes a theoretical argument why this correlation is to be expected. Third, other aspects of this theoretical argument are found to be reasonable.

Neuroimaging seems to be one of the present hypes in science and I suspect that large parts of it do not fare well even in light of the trivial picture of science presented above. Large parts of the theoretical part of explanation seem to be substituted by glossy and colorful pictures of the brain. Worse, a recent paper by Vul et al is arguing that even the correlations are statistically impossible or false. No correlation + no theoretical argument = no science.

(via)

Advertisement