Dear Marco,
It depends on the nature of the variability. If the variability is in source location then it might not be a good strategy to do statistical analysis at the source level, but you should rather do sensor statistics and then e.g. fit a dipole to your grand-averaged contrast waveforms.
If the variability is just in the strength of the effect then you can try group inversion and if that doesn't work you can try also individual inversions. Note that if you showed the significance of your effect by statistical analysis at the sensor level and used that to define your analysis window it doesn't make much sense to require whole-brain corrected FWE significance because your analysis is biased anyway. So you could just report the locations of peaks in your statistical image or use uncorrected threshold. Just explain in your write-up that this is to get an idea about sources rather than test for presence of the effect.
Best,
Vladimir
On 26 Sep 2012, at 09:11, Marco Buiatti <[log in to unmask]> wrote:
> Thanks Vladimir for the rapid response.
>
> A related question:
>
> I have 20 subjects, 64 channels, no single subject MRI and no
> anatomical landmark (no digitization). I am looking for a quite subtle
> effect which is significant at the sensor level (Fieldtrip non
> parametric statistical analysis) but variable from subject to subject.
> Due to this variability and the absence of single subject anatomical
> landmarks, I thought that I could do source reconstruction only on the
> grand-averaged ERP.
>
> Does it make sense with this variability to do source reconstruction
> on single subjects, or maybe do group inversion? Would this option
> account better for covariance variability across subjects?
>
> Thanks for your advice,
>
> Best,
>
> Marco
>
> On 25 September 2012 21:02, Vladimir Litvak <[log in to unmask]> wrote:
>> Dear Marco,
>>
>> It would be better to use normal or fine mesh if your computer power allows
>> it. I think these days there is no reason to use coarse, just when SPM8
>> first came out there were still many old an weak machines around. Hanning
>> window indeed multiplies the response in time and suppresses the edges. If
>> the interesting part of the response is in the middle of your time window it
>> would be advisable to use it.
>>
>> Best,
>>
>> Vladimir
>>
>>
>> On 25 Sep 2012, at 16:16, Marco Buiatti wrote:
>>
>>> Dear SPM masters,
>>>
>>> I am trying to perform source reconstruction on the grand-averaged
>>> ERPs from an EEG study (BrainAmp, 64 channels) with 20 subjects. Since
>>> the EEG electrode positions were not digitized, I am using the
>>> standard EEG template with standard electrode positions.
>>>
>>> My problem is that I see a quite wide variability in the sources by
>>> varying two parameters of the reconstruction:
>>> 1) use of coarse or normal cortical mesh
>>> 2) use or not of hanning window when inverting
>>>
>>> My questions are:
>>> 1) Is it normal to see large variability when using normal or coarse
>>> mesh? What would be the best for 64 electrodes?
>>> 2) What does the hanning window exactly refers to? I intuitively think
>>> that data in time are weighted by this window, but I could not find
>>> the exact info on the manual.
>>> 3) Any other important factor that you think is crucial for source
>>> reconstruction in this case?
>>>
>>> Thanks for your feedback,
>>>
>>> Best,
>>>
>>> Marco
>>>
>>> --
>>> Marco Buiatti, PhD
>>>
>>> CEA/DSV/I2BM / NeuroSpin
>>> INSERM U992 - Cognitive Neuroimaging Unit
>>> Bât 145 - Point Courrier 156
>>> Gif sur Yvette F-91191 FRANCE
>>> Ph: +33(0)169.08.65.21
>>> Fax: +33(0)169.08.79.73
>>> E-mail: [log in to unmask]
>>> http://www.unicog.org/pm/pmwiki.php/Main/MarcoBuiatti
>>>
>>> ***********************************************
>>
>>
>
>
>
> --
> Marco Buiatti, PhD
>
> CEA/DSV/I2BM / NeuroSpin
> INSERM U992 - Cognitive Neuroimaging Unit
> Bât 145 - Point Courrier 156
> Gif sur Yvette F-91191 FRANCE
> Ph: +33(0)169.08.65.21
> Fax: +33(0)169.08.79.73
> E-mail: [log in to unmask]
> http://www.unicog.org/pm/pmwiki.php/Main/MarcoBuiatti
>
> ***********************************************
|