Hi Alexander,
Thanks for the reply.
Interestingly, options #1 (SVC) and #3 (pre-masking) produce slightly
different statistics. This may be due to the resel formulation that the
SVC toolbox uses. Option #3 has the advantage of outputting a corrected
activation map in the masked region.
Thanks,
Gaby
On Wed, 12 Jan 2005 10:10:39 -0000, Hammers, Alexander
<[log in to unmask]> wrote:
>Dear Gaby,
>
>
>I'll let the stats gurus confirm but in principle if your analysis leads
to you looking at the same volume the stats should be the same.
>
>While I'm all for hypothesis driven research and looking at small
volumes, I would personally still prefer your first option. This is
because looking at the full SPM may give you more information. For
example, you get a feel for edge artefacts in comparisons which may not be
apparent when checking normalisation. In addition, in epilepsy but I guess
in other patient populations as well, nonsignificant blobbies all over the
place may be more prevalent than in control/control comparisons - if we
all stare hard enough for long enough, it all may become a pattern and
make sense. In summary, while the SVC toolbox will keep your stats neat,
why throw away all the other information?
>
>For an example of how we've gone about the problem in TLE, you could
check Hammers A et al. Brain 2002.
>
>All the best,
>
>Alexander
>----------------------------
>Dr Alexander Hammers, MD PhD
>
>MRC Clinician Scientist Fellow
>Honorary Clinical Lecturer
>Clinical Sciences Centre
>Room 243, MRC Cyclotron Building
>Faculty of Medicine, Imperial College London
>Hammersmith Hospital, DuCane Road, London W12 0NN
>
>and
>
>Honorary Lecturer in Neurology
>Department of Clinical and Experimental Epilepsy
>National Hospital for Neurology and Neurosurgery/ Institute of Neurology,
>UCL
>33 Queen Square, London WC1N 3BG
>
>Telephone +44-(0)20-8383-3162 (ext./direct line -3704 or -2651)
>Fax +44-(0)20-8383-1783 /-2029
>Email [log in to unmask], [log in to unmask]
>
>
>
>-----Original Message-----
>From: SPM (Statistical Parametric Mapping) [mailto:[log in to unmask]]On
>Behalf Of Gaby Pell
>Sent: 12 January 2005 09:23
>To: [log in to unmask]
>Subject: Re: [SPM] small volume correction
>
>
>Dear SPM users,
>
>What is the statistical difference between the following three approaches
>for small volume correction:
>(1) use of the SVC toolbox
>(2) explicit masking in SPM with the volume of interest
>(3) (even simpler) masking the data set with the VOI before implementing a
>standard SPM analysis
>
>Thanks,
>
>Gaby
>
>***************************************
>Gaby Pell, PhD
>Brain Research Institute
>Ground Floor, Neurosciences Building
>Austin & Repatriation Medical Centre
>Banksia Street
>West Heidelberg,
>Melbourne,
>Victoria, 3081
>Australia
|