Dear Alexa,

I explored a little bit yet still not sure what exactly you mean by
> A simpler and more usual approach is to test a region x condition  interaction. This is normally done by extracting the parameter estimates from a voxel or region of interest..

Is this parameter estimate the 90% C.I., or the signal change %, e.g.,
calculated from Marsbar?

> For laterality comparisons it is relatively unproblematic, as it is assumed that neural-BOLD coupling is equivalent.

Does this mean that I can only compare the left v.s. right side of a
same group (e.g., from elderly group), but not between groups (of
course the neural-BOLD coupling is equivalent unless there is evidence
for vascular disease like stroke). I guess you mean compare the
calculated laterality between groups.
 Sorry that I don't know how to calculate the laterality in fMRI..
Again using Marsbar and compare signal changes between two hemisphere?
BTW, I don't really know how to use script in Matlab for SPM.

Thank you all very much for nice suggestions!

Best regards,
Gao



This is normally done by extracting the parameter estimates from a
voxel or region of interest and performing an ANOVA outside SPM.


On 7 May 2012 23:08, Alexa Morcom <[log in to unmask]> wrote:
> Dear Gao
>
> You are right to observe that apparent laterality differences (or other
> differences in regional distribution of activity) between age groups (or
> indeed between other participant or drug treatment groups, or tasks) may
> reflect thresholding artefacts.
>
> The approach you mention might support this possibility, although it might
> be difficult to motivate the assumption that the 'top 1000' voxels are
> comparable between groups. You might also inspect unthresholded SPMs.
>
> A simpler and more usual approach is to test a region x condition
> interaction. This is normally done by extracting the parameter estimates
> from a voxel or region of interest and performing an ANOVA outside SPM. For
> laterality comparisons it is relatively unproblematic, as it is assumed that
> neural-BOLD coupling is equivalent.
>
> Good luck!
>
> Alexa
>
>
> On 5 May 2012 02:18, Gao Junling <[log in to unmask]> wrote:
>>
>> Dear SPM Experts,
>>
>> Here is a simple yet interesting question.
>> When we do SPM result, we usually set p = 0. 001 (uncorrected) or p = 0.05
>> (FWE correct), then we get the statistical activation mapping and how many
>> voxels activated. My questions is, can we set a specific number of activated
>> voxels (e.g., 1000 voxel totally) and then get the activation map?
>>
>> The reason to do this is that I have two groups of subjects, young and
>> old, although the final activation map is different between young and old.
>> However, after setting a lower threshold for young subjects, I found the
>> activation maps between the two groups are rather similar. This makes me
>> wonder whether I can actually set a total voxel number and see how much
>> similar the maps of the two group will be. My feeling is that the so-called
>> HAROLD model(Hemispheric asymmetry reduction in older adults) might not be
>> an absolute distinction, but a matter of levels of activations.
>>
>> It seems SPMS can only set p-level before you can get total activated
>> voxels. Anyone here know about the other way? Thank you very much!
>>
>> Best wishes,
>> Gao
>>
>
>
>
> --
> Dr. Alexa Morcom
>
> Centre for Cognitive & Neural Systems
> Centre for Cognitive Ageing & Cognitive Epidemiology
> Psychology, University of Edinburgh
> http://www.ccns.sbms.mvm.ed.ac.uk/people/academic/morcom.html
>
> The University of Edinburgh is a charitable body, registered in Scotland,
> with registration number SC005336
>
>
>
> The University of Edinburgh is a charitable body, registered in
> Scotland, with registration number SC005336.
>