Dear Morgan:
Smoothing less at an individual level allows one greater precision in the
anatomical specification. Smoothing more is often necessary for group analysis
because of differences in anatomy. The larger smoothing kernel somewhat
compensates for variability in the normalization processs. This is not any issue
specific to ranfx per se rather to group analyses. The one caveat for smoothing
a second time is the problem with losing a layer of edge voxels because of
blurring between brain and the points outside the brain (which SPM codes as
NaN's.) This has been discussed on the list before. See posts by R. Poldrack
some time back.
Regards,
Darren
Morgan Willson wrote:
> Dear Bryan,
>
> Why smooth the original images by only 4mm in the first place why not smooth
> them to 8mm to avoid having to re-smooth them in order to do the second
> level analysis. Is there some benefit to the smaller smoothing kernal for
> individual analysis, or conversly a benefit to a larger smoothing window for
> group analysis?
>
> Morgan Willson
> Biomedical Engineering
> University of Alberta
> Canada
> [log in to unmask]
>
> ----- Original Message -----
> From: "Bryan Strange" <[log in to unmask]>
> To: <[log in to unmask]>
> Sent: Monday, September 24, 2001 12:01 PM
> Subject: Re: second smoothing
>
> > Dear Tony,
> >
> >
> > > I have smoothed my normalised images to 4mm and looked at data for
> > > individuals. Now I wish to go through to a second level, random effects,
> > > analysis.
> > > 1) Can I simply smooth the con* images before entering them into the
> > second level?
> >
> > Yes
> >
> > > 2) Suppose it is OK to smooth the con* images by X mm. Is it true that
> > > the 'total' smoothing would be the root of 4*4 + X*X ?
> >
> > Yes
> >
> > > 3) Finally, what is the current consensus on the appropriate amount
> > > of 'total' smoothing, when the areas of interest are in DLPFC?
> >
> > No hard and fast rule on this but 8 or 10 mm 'total smoothing' seems to
> work well.
> >
> >
> > Phillipe - Tony's motivation for further smoothing is that his data are
> not sufficiently smoothed for a RFX analysis. Smoothing the con images
> avoids re-smoothing normalised functional images and re-estimating the
> first-level model.
> >
> > Best regards,
> >
> > Bryan
> >
> > Bryan Strange
> > Wellcome Dept Cognitive Neurology
> > Institute of Neurology
> > 12 Queen Square
> > London WC1N 3BG
> > +44 (0) 207 833 7485
> > [log in to unmask]
> >
> > and
> >
> > Institute of Cognitive Neuroscience
> > University College London
> > 17 Queen Square
> > London WC1N 3AR
> > +44 (0) 207 679 5431
--
----------------------------------------------------------------------
Darren R. Gitelman, M.D. E-mail: [log in to unmask]
Cognitive Neurology and Voice: (312) 908-9023
the Alzheimer¹s Disease Center Fax: (312) 908-8789
http://www.brain.nwu.edu
Northwestern Univ., 320 E. Superior St., Searle 11-470, Chicago, IL 60611
----------------------------------------------------------------------
|