Dspm and loose orientation solution

Hi All,
  I have a question relating modeling the null distribution in a dspm map with an inverse operator that is not obtained using --fixed.
    When mne_make_movie is run with --spm option but without the --signed option which of the following is done?
1. Noise normalize each component first and then do sqrt(sum of squared x,y,z) or
2. Do sqrt(sum of squared x,y,z) then noise normalize

Thanks,
Hari

Hi Hari,

to me the noise normalization is location specific and not orientation specific.
It means that 1 and 2 are equivalent.
Regarding the --signed option it should not have any influence.

Alex

Hi Alex,
   Thanks for the response. But the estimated source noise variance is
different for the 3 components, right? So why is it that we only do a
location specific normalization? Is it because all we want to do is
reduce point spread and localization bias at the vertex level?

Thanks,
Hari

? Thanks for the response. But the estimated source noise variance is
different for the 3 components, right?

correct

So why is it that we only do a
location specific normalization?

because you pull the variance for all 3 components to have 1 value
that is then used to normalize the 3 components at each location.

Is it because all we want to do is
reduce point spread and localization bias at the vertex level?

You can see things like this.

Alex