# Noise covariance from emptyroom data

**URL:** <https://mne.discourse.group/t/noise-covariance-from-emptyroom-data/139>\
**Category:** Mailing List Archive (read-only)\
**Tags:** list-archive\
**Created:** [June 23, 2009, 5:58pm UTC](https://mne.discourse.group/t/noise-covariance-from-emptyroom-data/139 "2009-06-23T17:58:17Z")\
**Posts on this page:** 1\
**Page:** 1

<div class="post-metadata">

**Author:** ![system](https://global.discourse-cdn.com/free1/uploads/mne/original/1X/85cc6bd2b69cb698a166dc6d880fb550510d0144.jpeg) [@system](https://mne.discourse.group/u/system)\
**Post date:** [June 23, 2009, 5:58pm UTC](https://mne.discourse.group/t/noise-covariance-from-emptyroom-data/139/1 "2009-06-23T17:58:17Z")

</div>

Hello,

Does anybody have thoughts on what kinds of preprocessing the emptyroom  
data must undergo before computing the noise covariance matrix for source  
modeling?

An obvious suggestion is to subject the emptyroom data to identical  
preprocessing steps (SSS with identical parameters, or some projection,  
before source modeling). How far can we take this? For instance, if one  
wants to estimate oscillatory activity at the cortical level for a narrow  
freq. band (which has very different noise characteristics from the entire  
band of the recording), should we compute the noise covariance from  
filtered emptyroom data?

More generally, if one is analyzing several freq. bands, is it worthwhile  
estimating the noise covariance (and thus, a separate inverse operator)  
for each frequency band?

Regards,  
Pavan
