> There are also issues about the validity of comparisons between single subjects
> and populations in terms of how normally distributed the data are. Unlike
> the case of a comparison between populations, the statistical tests are
> particularly susceptible to violations of the assumptions about normality.
> For this reason, it is necessary to smooth the data more to achieve suitably
> normally distributed errors. From the information I have available, I believe
> that 12mm smoothing may be enough.
Alternatively, you could use non-parametric statistics, which does not assume
a normal distribution for the errors. However, to get a p value less than 0.05
you would need at least 20 (or 21???) subjects in the analysis (as there are
only 20 ways of permuting the data). See: http://www.fil.ion.ucl.ac.uk/spm/snpm/
Best regards,
-John
--
Dr John Ashburner.
Functional Imaging Lab., 12 Queen Square, London WC1N 3BG, UK.
tel: +44 (0)20 78337491 or +44 (0)20 78373611 x4381
fax: +44 (0)20 78131420 http://www.fil.ion.ucl.ac.uk/~john
|