[Mne_analysis] Covariance and noise density in simulation

[Alexandre Gramfort]

        External Email - Use Caution        

hi Taylor,

the noise cov in MNE simulation module is not frequency specific. We would
need to add support for Multivariate autoregressive model (MVAR) for noise
simulations to do what you want. Any help to allow us to achieve more
realistic simulations is very welcome and helpful for many people.

Alex

On Tue, Aug 13, 2019 at 10:53 PM Taylor Williams
<williams.taylor at gmail.com> wrote:
>
>         External Email - Use Caution
>
> Hello,
>
> I have some questions about how covariance in MNE simulation methods corresponds to noise density. I'm seeking to construct an ad hoc covariance matrix that corresponds to a particular noise density at a particular sampling frequency.
>
> For example, I'd like to simulate evoked data for an MEG array that has 3 fT/sqrt(Hz) white noise density at 600 Hz sampling frequency. Documentation for create_ad_hoc_cov() states that the default values for magnetometers is 20 fT.
>
> To properly simulate 3 fT/sqrt(Hz) at 600 Hz, should I calculate the total noise as 3*sqrt(600) = 73.4 fT and scale my cov matrix from 20 fT accordingly, or do the MNE methods account for system bandwidth in simulation? Is it common to simulate the full noise bandwidth then apply filtering to keep the frequency band of interest?
>
> Thanks,
> Taylor Williams
>
>
>
>
>
>
> _______________________________________________
> Mne_analysis mailing list
> Mne_analysis at nmr.mgh.harvard.edu
> https://mail.nmr.mgh.harvard.edu/mailman/listinfo/mne_analysis