Dear Tugan,
Regarding the model for signal used in that report, it is worth
reiterating that it was a distributed signal having the properties of
a Gaussian random field itself. This is, of course, something you
might already have known.
Sincerely,
Eric
> Dear Tugan,
>
> > I want to apply the statistical power analysis described in "Detecting
> > activations in PET and fMRI:levels of inference and power", Neuroimage 40,
> > 223-235 (1996). In page 227, under the topic "A Power Analysis" it is said
> > that "The shape of the signal is characterized by the width (f) of the foci
> > expressed as a proportion of W". I cannot clearly understand what is meant
> > by "the signal" and how we get this f value or what it corresponds to. I'm
> > not a statistical expert and I'll appreciate any help on this.
>
> In this particular instance, f is the width of signal sources generated
> in some simulations which aimed at measuring the characteristics (e.g.
> sensitivity) of a given statistical analysis and to draw useful
> conclusions for the analysis of real fMRI or PET data.
>
> That means f was explicitly known and used to generate simulated data.
> Since the smoothness W (e.g. 3 voxels in all three directions) can be
> seen as some measure of the resolution within a statistical volume,
> choosing f as a proportion of W (e.g. f = 0.5, i.e. width is 1.5 voxels)
> is useful to compare signal widths directly to the smoothness.
>
> Of course, in real data one doesn't know f and W and one has to estimate
> these values somehow...
>
> Best wishes, Stefan
> --
>
>
> Stefan J. Kiebel
> Dept. of Neurology
> Friedrich Schiller University
> Philosophenweg 3
> 07740 JENA
> Germany
>
> Tel. : +49 3641/9-35285
> FAX : -35399
> email: [log in to unmask]
> WWW : www.neuro.uni-jena.de
>
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
|