Dear Hernando:
Resel computation is done after the statistics estimation stage. In fact, as far
as I know, smoothness estimation and resel computation are performed in the
residual images (see spm_spm.m) as opposed to the actual statistics images. This
is done to avoid bias by physiological signal. See
@Article{kiebel1999,
author = "S. J. Kiebel and J.-B. Poline and K. J. Friston and A. P.
Holmes and K. J. Worsley",
title = "Robust smoothness estimation in statistical parametric maps
using standardized residuals from the
general linear model",
journal = "NeuroImage",
year = "1999",
volume = "10",
pages = "756-766"
}
Any changes in the analysis procedure that occur upstream w.r.t. model
estimation are thus expected to lead to changes in resel values.
Hope this answers your question...
-Alexandre
#--------------------------------------------------------#
| Alexandre Andrade |
| Service Hospitalier Frederic Joliot, CEA/DRM |
| 4, Place du General Leclerc |
| 91401 Orsay cedex, France |
| Phone: +33 1 69 86 78 21 |
| Fax: +33 1 69 86 77 86 |
| e-mail: [log in to unmask] |
#--------------------------------------------------------#
> Dear SPM experts,
>
> We wanted to compare the results when using ANCOVA vs. Proportional
> Scaling for global mean signal correction. We noted that the two methods
> gave different resel sizes. In all of our comparisons so far, the ANCOVA
> method gave a resel unit with fewer number of voxels than the Proportional
> Scaling.
>
> My understanding was that the resel value is determined at the image
> smoothing stage. Thus, regardless of any statistics that follow, the resel
> unit should not change.
>
> I would appreciate any ideas and explanation on this matter.
>
> Thanks,
>
> Hernando Ombao
>
> ----------------------------------------------------------------------------
> Hernando Ombao, Ph.D.
> Assistant Professor of Statistics and Psychiatry
> University of Pittsburgh
>
> Department of Statistics Phone: 412-624-8719
> 2706 Cathedral of Learning Fax : 412-648-8814
> University of Pittsburgh Email: [log in to unmask]
> Pittsburgh, PA 15260 Webpage:www.stat.pitt.edu/~ombao
> ----------------------------------------------------------------------------
|