Tuesday, October 11, 2011

Erroneous analyses of interactions in neuroscience: a problem of significance

http://www.nature.com/neuro/journal/v14/n9/full/nn.2886.html

Whatever the reasons for the error, its ubiquity and potential effect suggest that researchers and reviewers should be more aware that the difference between significant and not significant is not itself necessarily significant.

A fictive example would be “Hippocampal firing synchrony correlated with memory performance in the placebo condition (r = 0.43, P = 0.01), but not in the drug condition (r = 0.19, P = 0.21)”. When making a comparison between two correlations, researchers should directly contrast the two correlations using an appropriate statistical method.

No comments: