Hi Rachel,
for the CheckReg option you would overlay your 2nd level blobs onto a
suitable image (e.g. canonical/single_subject_T1, a normalised fMRI
image or something else), and then extract data from all
normalised&smoothed fMRI images of all subjects and sessions. This is not
exactly the same data as SPM would use for estimation and plots on 1st
level, because scaling and high-pass filtering are not applied and data is
not adjusted for any specific contrast.
As I already said, since the tool will extract data at each voxel in your
blobs for all subjects and timepoints (it does not know about onsets of
your events), you will need quite a lot of memory (or do extraction for
each subject separately).
Hope this helps,
Volkmar
On Thu, 15 Feb 2007, Rachel Kozink wrote:
> Hi Volkmar,
>
> Thanks for the quick response. If I am unable to see what the waveforms
> are that are driving my results, then how do I know what's happening? Are
> there other ways of attributing my results to an increase or decrease in
> either condition a or b between my two groups of subjects? If I were to
> use the checkreg option, are you referring to my con images that I create
> for each of my subjects? I am asking these questions because we have some
> in-house software that presents the raw data to me for a specified epoch
> around my events, and I am not seeing the same results in the same areas
> as when I use SPM. One downfall of the in-house program, however, is that
> it is only using the two time points prior to the event as a baseline
> correction for the signal. However, I would think the shape should be
> relatively similar.
>
> I apologize if my questions are somewhat naive, but this is really the
> first large analysis that I have done with SPM and want to make sure I
> understand what the results mean.
>
> Thanks again,
> Rachel
>
>
>
>
> Volkmar Glauche <[log in to unmask]>
> Sent by: "SPM (Statistical Parametric Mapping)" <[log in to unmask]>
> 02/15/2007 09:35 AM
> Please respond to
> Volkmar Glauche <[log in to unmask]>
>
>
> To
> [log in to unmask]
> cc
>
> Subject
> Re: [SPM] random effects plotting
>
>
>
>
>
>
> Hi Rachel, Cedric et al,
>
> afaik something like this is not implemented in SPM. A workaround for
> plotting would be to run an additional fixed-effects analysis over all
> subjects and sessions, but this won't work for large #scans and #subjects.
>
> Another way to go might be the "Extract&Plot Data" plugin from CheckReg
> context menu (included in SPM5, available from SPM extensions page for
> SPM2). You would overlay your results as blobs in CheckReg and then
> Extract&Plot data for all raw images from all subjects. However this will
> probably stress your computer memory quite a lot, and you will have to do
> any averaging over sessions/subjects by hand.
>
> Volkmar
>
> On Thu, 15 Feb 2007, Cédric Lemogne wrote:
>
>> Dear all,
>> I can not answer to Rachel but I would be very happy if any advanced
> user could ! ;-)
>> (I am using SPM 5)
>> Many Thx
>> Cédric
>>> Message du 15/02/07 15:00
>>> De : "Rachel Kozink"
>>> A : [log in to unmask]
>>> Copie à :
>>> Objet : [SPM] random effects plotting
>>>
>>> I have 15 subjects, split into two groups: group 1 (7 subjects) and
> group 2
>>> (8 subjects). Each subject performed 9 runs of the same task, where we
> are
>>> interested in cond a > cond b across all 9 runs. I modelled each
> subject
>>> individually at the first level, then took the contrast images (a>b) to
> the
>>> second level, random effects analysis. I performed a two sample t-test
> and
>>> got some interesting results. Is there any way that I can look at
>>> the "raw" data that is driving those effects (i.e. separate waveforms
> for
>>> cond a group1, cond b group 1, cond a group2 cond b group2) in each of
> my
>>> activated clusters? I greatly appreciate any help. I am using spm2.
>>>
>>> Rachel
>>>
>>>
>
>
--
Volkmar Glauche
-
Department of Neurology [log in to unmask]
Universitaetsklinikum Freiburg Phone 49(0)761-270-5331
Breisacher Str. 64 Fax 49(0)761-270-5416
79106 Freiburg http://fbi.uniklinik-freiburg.de/
|