Data conversion to reduce multiplicative uncertainty and its application to analysis of variance

We propose a method to reduce inter-subject variation which is due to multiplicative factor assumed to be present in measurement of MEG amplitudes. The method is to multiply each subject's data by a constant such that inter-subject variance is minimized. After the conversion, usual ANOVA is formally applied using newly derived ‘F’-tables. Simulation shows that when the subject effect is indeed multiplicative, the proposed conversion yields a stronger power of detection of the effect of experimental conditions than the conventional ANOVA.