Dear SPMers,
I have read with great interest the discussion about the way fMRI (or
generally speaking imaging or even behavioral results) should be presented.
I may have missed a few steps in the discussion, so I hope the point has
not been raised already in the way I would like to develop it, but it seems
to me that the underlying question is "Do we believe in stats or in our
eyes?" And of course the answer should be both.
On the one hand, stats are our common law, that is we generally agree that
only results that reach a given statistical threshold are supposed to be
robust enough to deserve discussion and interpretation, otherwise we
increase the risk to interpret simple noisy patterns.
On the other hand, eye inspection on continuous maps may indeed nicely show
that the brain region found activated at p < .XX may not be the only brain
area that COULD participate in the component of interest. In my (probably
naive) opinion however, there is a potential problem with the "COULD
participate", because it becomes rather subjective (or let's say more than
usual). Hence I would be happy having a better constraint on this "COULD
participate" probability. It seems to me that this can be achieved using
bayesian posterior probability maps, in which at least we try to quantify
this chance. So what do you think of continous PPM's ?
Anyway, thanks to all of you for this very stimulating discussion. The
field needs it.
-- Philippe Peigneux
|