Li, Yu wrote:
> Dear SPMers,
>
> Need help in calculating sample size from activation maps data in fMRI study.
> How do you calculate how many subjects are needed for a particular p -value and also for corrected and uncorrected significance values .
> We know that sample size of 12 - 18 subjects is needed for random effects analysis but is there anything more specific for a
> grant submission.
This calculation requires that you know (i) the size of the effect you wish to measure (ie the beta
value from an SPM analysis or value in a con image) and (ii) the variability of the size of that
effect. The only way to find these out is to do a pilot study (or maybe these
will have been reported in the literature by other researchers). Then you plug these
values into standard formulae to calculate numbers of subjects required to achieve a given
sensitivity (power) at a particular level of specificity.
The problem with neuroimaging, however, is that you'd need to do this
for all the regions of the brain that you are interested in
finding effects in. And because there may be so many this effort
is substantial.
For more on this I can recommend the landmark paper:
Desmond and Glover (2002) Estimating sample size in functional MRI: statistical
power analyses. Journal of neuroscience methods, 118, 115-128.
Best wishes,
Will.
>
> Thanks
>
>
>
> Thanks
> Julia Li
>
>
>
--
William D. Penny
Wellcome Department of Imaging Neuroscience
University College London
12 Queen Square
London WC1N 3BG
Tel: 020 7833 7478
FAX: 020 7813 1420
Email: [log in to unmask]
URL: http://www.fil.ion.ucl.ac.uk/~wpenny/
|