Faced with a correlation between two variables — chickens and eggs, say — how do you know which is causing the other? Questions such as this, common to fields as disparate as climatology and physiology, are typically unravelled with a statistical technique called Granger causality. But Guido Nolte of the Fraunhofer FIRST institute in Berlin, Germany, and his co-workers say that this method can falsely attribute causality.
They describe a new technique that relies on how phase differences between the driving and dependent variables change with the frequency of their fluctuations. This proves more reliable when the data contain a lot of noise. The authors use their discovery to identify the order in which certain brain regions stimulate others when human subjects shift from a relaxed to an alert state.
Rights and permissions
About this article
Cite this article
Statistics: Who's the driver?. Nature 453, 1146 (2008). https://doi.org/10.1038/4531146b
Published:
Issue Date:
DOI: https://doi.org/10.1038/4531146b