Confounded correlations: statistical limitations in the analysis of interregional relationships of cerebral metabolic activity.

Various statistical approaches have been applied to tomographic determinations of glucose utilization putatively to gain insight into functional interrelationships in the brain. An approach based on the use of a partial correlation coefficient has been proposed to eliminate the distorting influence of systematic interpatient differences in glucose utilization. It is shown, by examination of explicit statistical models, that this form of analysis does not permit unambiguous conclusions to be drawn.