---- Original Message -----
From: Matthew Brett <[log in to unmask]>
Date: Monday, February 28, 2005 1:49 pm
Subject: Re: [SPM] Any Papers on Presenting fMRI Results?
> Hi,
>
> > On the other hand, this sounds like paying very little attention to
> > statistics, doesn't it? Transposing this approach to behavioral
> data,> it sounds to me like simply looking at histograms and saying
> "Well,> guys, this bar seems a little bit different/similar from
> that one, so
> > let's say that there is/isn't a difference".
>
> Thanks for bringing this up - this issue is really central to the
> argument. The comparison to behavioral data is apt, because what you
> are suggesting is that we simply do not show the (e.g) reaction times
> for tests that did not pass the significance level. Let's say we have
> two tasks A and B, compared to control. A passes significance at
> p=0.05, B doesn't p=0.04. It could very easily be that B has even has
> a higher effect size than A. It seems to me very misleading to report
> 'A is significant' without 'B is very close to A'. The continuous map
> provides this information in a rather compact way.
I agree that continuous maps are a nice way of presenting results,
especially with the significant areas demarcated as in your example. A
further question is what should we be graphing in such maps? In many
cases people graph t or z stats, but perhaps it might be preferable to
map % signal change or another similar measure that gives an idea of
average magnitude of within-subject signal change (thresholded of course
by the appropriate statistical parameter or statistically-defined
cluster boundary). This would be more akin to behavioral data
presentation (although of course the relative nature of fMRI makes the
analogy somewhat incorrect).
Tom
University of Wisconsin-Madison
|