Dear Vladimir

I have a question regarding my approach. I did my source localisation on the grand average ERP data of two conditions across all the subjects and derived the source waveforms.

Then instead of constraining the sources on individual subjects , i applied it on on Jackknife averages (Jackknife is a procedure in which from the grand average, one subject is deducted in every iteration) and obtained the jackknife average source waveform.

Now i am applying the LFP model on these jackknife average source waveform to decide the winning model and the connectivity changes thereby between the two conditions. Is this approach suitable from DCM point of view for correct estimation.

In addition, i would also like to know about the meaning of positive and negative values of mEp and if they have any units.

regards

Rajan Kashyap

Hong Kong Baptist University

On Sat, Apr 18, 2015 at 2:26 AM, Vladimir Litvak <[log in to unmask]> wrote:

Dear Rajan,

You can look at the video talk at http://www.fil.ion.ucl.ac.uk/spm/course/video/#MEEG_BMS for the relevant theory. The practicalities are described in Chapter 37 of SPM12 manual.

Best,

Vladimir

On Fri, Apr 17, 2015 at 1:31 PM, Rajan Kashyap <[log in to unmask]> wrote:
Dear Vladimir

I have another question. It may be very basic, but important for my analysis.
I have 21 subjects. Each subject has 2 models (1)Forward and (2) Forward backward (FB) model.
Single subject comparison shows FB model as winning model and the same exists for 18 subjects.

I am unable to understand how to proceed for group level inference on model selection.

Kindly advice.

Regards
Rajan Kashyap
Hong Kong Baptist University

On Fri, Apr 17, 2015 at 12:44 AM, Vladimir Litvak <[log in to unmask]> wrote:
Dear Rajan,

There are 3 different choices in the GUI. One is where you choose the analysis type ERP/CSD/IND etc. There you should choose ERP. Second is where you choose the neural model ERP/SEP/CMC etc. You can start with ERP there as well. See http://journal.frontiersin.org/article/10.3389/fncom.2013.00057/full for more detailed discussion.

Finally in the spatial model part ECD/IMG/LFP you should choose LFP and refer to the chapter for how to specify channels and source names for DCM.

Best,

Vladimir

On Thu, Apr 16, 2015 at 6:50 PM, Rajan Kashyap <[log in to unmask]> wrote:
Dear Vladimir
I am just confused in that part. Instead of using the SSR, Can i Use ERP as spatial model ??.
Then i can use ERP as my neuronal model.

regards
Rajan

On Thu, Apr 16, 2015 at 8:27 PM, Vladimir Litvak <[log in to unmask]> wrote:
Dear Rajan,

You should use the LFP spatial model for source data. The neural model can be any of ERP, CMC or NMM. See the rat anaesthesia tutorial chapter in the manual for an example of how to work with source level data.

Best,

Vladimir

On Thu, Apr 16, 2015 at 1:36 PM, Rajan Kashyap <[log in to unmask]> wrote:
Dear all
I want to use DCM on the source waveforms of the face ERP data.
Especially i want to see how each hemisphere works.

I want to model the source waveforms using spatial model as ERP and connectivity model also as ERP to see what are the differences in connection parameters in the two conditions.
Can i do it??

Regards
Rajan Kashyap