According to Gareth, EBB also has a tendency to at least bias the locations to be the same across conditions so it's not unreasonable.

Vladimir

On Wed, Aug 26, 2015 at 6:27 PM, Jane Kouptsova <[log in to unmask]> wrote:

Vladimir, thank you for your quick response. I am using EBB - is this assumption also true here, or would my results be unusual for this algorithm?

Thanks,

Jane

On Wed, Aug 26, 2015 at 12:29 PM, Vladimir Litvak <[log in to unmask]> wrote:

Dear Jane,

If you are using the MSP method (GS or ARD options) running it on the three conditions combined this makes a lot of sense as this method assumes that the same sources are active in all the conditions, just their level of activation differs. So the algorithm will try to find sources that can fit all the three conditions. If you select e.g. the 'IID' option, this should not be the case.

Best,

Vladimir

On Wed, Aug 26, 2015 at 3:47 PM, Jane Kouptsova <[log in to unmask]> wrote:

Dear all,

I am running an MEG analysis using SPM12, where I look at differences between three conditions, say 1, 2 and 3. When I run contrasts, I expect 3vs1 and 3vs2 to give similar results, and this is what I'm getting, however, more than being similar, many of the cluster peak coordinates are identical. Is this a cause for concern?

The Z, D-uncorr and other values from the contrast are slightly different even when peaks are identical, and running a 1vs2 contrast also gives some results. So I know I didn't accidentally assign the same set of trials to both conditions. Is there anything else I should check to be sure of the output?

Many thanks,

Jane