<div dir="ltr">Hi all,<div><br></div><div>I noticed that if I scale the noise covariance matrix, it does not change the source estimates in any way. In <a href="https://martinos.org/mne/stable/auto_tutorials/plot_mne_dspm_source_localization.html#sphx-glr-auto-tutorials-plot-mne-dspm-source-localization-py">this example</a> from the example gallery if I change  the method to &quot;MNE&quot; before the first stc calculation and then do this:</div><div><br></div><div><pre style="font-family:Consolas;font-size:10.8pt">inverse_operator = make_inverse_operator(info, fwd, noise_cov,<br>                                         <span style="color:rgb(102,0,153)">loose</span>=<span style="color:rgb(0,0,255)">0.2</span>, <span style="color:rgb(102,0,153)">depth</span>=<span style="color:rgb(0,0,255)">0.8</span>)<br>stc = apply_inverse(evoked, inverse_operator, lambda2,<br>                    <span style="color:rgb(102,0,153)">method</span>=method, <span style="color:rgb(102,0,153)">pick_ori</span>=<span style="color:rgb(0,0,128);font-weight:bold">None</span>)</pre><pre style="font-family:Consolas;font-size:10.8pt"><span style="color:rgb(0,0,128);font-weight:bold">from </span>copy <span style="color:rgb(0,0,128);font-weight:bold">import </span>deepcopy<br>noise_cov_scaled = deepcopy(noise_cov)<br>noise_cov_scaled[<span style="color:rgb(0,128,128);font-weight:bold">&#39;data&#39;</span>] *= <span style="color:rgb(0,0,255)">1000000<br></span>inverse_operator_scaled = make_inverse_operator(info, fwd, noise_cov_scaled,<br>                                         <span style="color:rgb(102,0,153)">loose</span>=<span style="color:rgb(0,0,255)">0.2</span>, <span style="color:rgb(102,0,153)">depth</span>=<span style="color:rgb(0,0,255)">0.8</span>)<br>stc_scaled = apply_inverse(evoked, inverse_operator_scaled, lambda2,<br>                           <span style="color:rgb(102,0,153)">method</span>=method, <span style="color:rgb(102,0,153)">pick_ori</span>=<span style="color:rgb(0,0,128);font-weight:bold">None</span>)</pre></div><div><br></div><div>Then stc_scaled contains exactly the same data as stc. With &quot;dSMP&quot; and &quot;sLORETA&quot; the result do differ. Why doesn&#39;t &quot;MNE&quot; care about the scale of the covariance matrix?<br></div><div><br></div><div>Evgenii</div></div>