Thanks for the lightning-speed reply!
And recommendations for single-case studies?
On 11/02/2016 12:28, Vladimir Litvak wrote:
> This is not surprising because with a smooth prior like IID or COH you
> always get some activation everywhere. If you want to get focal results
> you should proceed to statistical analysis across subjects (e.g. compare
> two conditions).
>
> Model evidence is indeed one way to choose the best method. We would
> recommend to use MSP with group inversion if you have multiple subjects
> but some people get better results with COH or IID.
>
> Best,
>
> Vladimir
>
> On Thu, Feb 11, 2016 at 12:52 PM, gj <[log in to unmask]
> <mailto:[log in to unmask]>> wrote:
>
> Hi,
>
> We did inverse reconstruction in SPM 12 using the COH/LORETA-like
> smooth prior inversion model.
>
> Like Jun Wang's 2012 post "EEG source reconstruction in wrong place"
> https://www.jiscmail.ac.uk/cgi-bin/webadmin?A2=spm;f26e48d5.1210
>
> this leads to source localization "everywhere", in our case, where
> ever there are gyri (when we "display" the "image" after "invert"),
> although it appears to show stronger activation where we expect.
> However, the source is localized to where we would expect it to be
> using GS/greedy search.
>
> Could this be a diagnostic of something wrong we have done? Maybe
> inappropriate to ask here, but any pointers to where we can find
> guidelines/recommendations as to how to choose what inversion model
> to use (try them all and choose the one with the highest log evidence?)?
>
> Many thanks in advance!
>
>
|