 |
 |
Dear Rebecca > I am filling out an ethics committee form which requests the > statistical power of the proposed experiment. I am doing a random > effects analysis on two groups of subjects with 12 patients compared > with 12 controls, according to the standard RFX recommendations. I > assume you did the appropriate power calculations on which to base > these recommendations about subject numbers for RFX analysis? If so, > and if the numbers are to hand, I want be very grateful to know the > results. Re: simple question - no its not! A power analysis requires a specification of the alternate hypothesis. There is no general specification in brain mapping (you would have to specify where the activation was, which modality you used, what constitutes an activation, whether the inference was voxel-, set- or cluster-based, whether the inference was corrected ... and so on). I would say, with an air of casual authority, that 'it is generally accepted that 12 subjects are sufficent to detect activations of 0.25% (fMRI) or 4% (PET) at a specificiy of 95% and sufficient sensitivity to detect at least one activated brain region' and then quote: Friston KJ Holmes A Poline J-B Price CJ Frith CD Detecting activations in PET and fMRI: levels of inference and power NeuroImage 1996;4:223-235 Friston KJ Holmes AP Worsley K. How Many subjects constitute a study? NeuroImage 10 1-5 (1999). The above statement in meaningless but they might enjoy the papers. I hope this helps - Karl %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%