[Mne_analysis] Statistical Thresholding of dSPM maps

[Daniel Goldenholz]

Alex

Three thoughts on this, take em or leave em.

If you wish to correct for multiple comparisons, you should also keep track
of the spatial correlation structure of the source space (implied by the
inverse model).

Similarly, there may exist some temporal correlation structure, however that
is less well defined and would require you to make some additional
assumptions about what you believe you are measuring.

Lastly, if you are considering temporal correlation structure, you would
need to be very careful regarding a entity called the "noise covariance
matrix" in MNE, which assumes that it contains a temporally stationary
random process. Many use "baseline brain activity" to produce the noise
covariance matrix, and it is possible that such data is not such a process.
I have been advocating using "empty room" measurements to produce a more
reasonable noise covariance matrix.

DG

On Mon, Sep 15, 2008 at 7:53 AM, Alex Clarke <alex at csl.psychol.cam.ac.uk>wrote:

> Hello,
>
> I had a few questions about thresholding dSPM maps. It was my understanding
> that the dSPM maps provide an F-value (as orientation was not constrained)
> that can be related to a p-value. However, I was wondering if there were any
> multiple comparisons procedures implemented in MNE?
>
> Thanks for your help,
>
> Alex
> _______________________________________________
> Mne_analysis mailing list
> Mne_analysis at nmr.mgh.harvard.edu
> https://mail.nmr.mgh.harvard.edu/mailman/listinfo/mne_analysis
>



-- 
Daniel Goldenholz MD, PhD
--------------------------------------------------------
http://www.nmr.mgh.harvard.edu/~daniel
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://mail.nmr.mgh.harvard.edu/pipermail/mne_analysis/attachments/20080915/7e0c1c4b/attachment.html