Alex,
> I have a question concerning the signal to noise ratio.
> We are planning a new experiment with old and young
> subjects to figure out the magnitude of activation in a
> specific region. In the case of a significant difference
> between both groups we would like to calculateor to
> determine the signal to noise ratio. To do this I think we
> need the signal variance and the noise variance of our two groups.
> But how can I get this values for each single subject of the old and
> young group using SPM99?
The noise variance can be obtained from the residual variance images;
if you have a single multisubject model, then ResMS is exactly what
you want, if you have sets of intrasubject analyses, you could average
all of the ResMS's together.
For the "signal variance" do you mean overall signal magnitude, or how
that signal varies between subjects? For the former, you can just use
the contrast images (make sure all of your models and contrasts have
been defined similarly), constructing one huge contrast if you fit a
single multisubject model, or averaging each subject's contrast image
if you fit different models. For the later, use the variance of the
intrasubject contrast images. The easiest way to get a variance of a
set of images is to use "Basic Stats": Fit a one-sample t-test model,
and then the resulting ResMS image will be the variance of the images.
Hope this helps.
-Tom
-- Thomas Nichols -------------------- Department of Biostatistics
http://www.sph.umich.edu/~nichols University of Michigan
[log in to unmask] 1420 Washington Heights
-------------------------------------- Ann Arbor, MI 48109-2029
|