Dear Gareth and Vladimir,
Great thanks for your kindly reply! I am now clear about the dipole model. Many thanks!
Bests,
Sun Delin
________________________________________
From: Vladimir Litvak [[log in to unmask]]
Sent: Thursday, January 03, 2013 11:44 PM
To: Barnes, Gareth
Cc: sundelin; [log in to unmask]
Subject: Re: [SPM] Variance explained for a dipole model of ERP data
Thanks Gareth, that explains it. I should have figured it out event
without looking at the code so sorry for wasting your time.
Vladimir
On Thu, Jan 3, 2013 at 3:41 PM, Barnes, Gareth <[log in to unmask]> wrote:
> Hi Sun
> The main difference is that Vladimir's script calculates the percentage variance over the whole time period of the epoch, whereas spm_eeg_inv_vbecd caculates percentage variance for the time point you chose to fit.
> When I adjusted Vladmir's script to look at a single time instant I got the same answer.
> best
> gareth
>
> ________________________________________
> From: sundelin [[log in to unmask]]
> Sent: 03 January 2013 13:49
> To: Barnes, Gareth
> Subject: RE: [SPM] Variance explained for a dipole model of ERP data
>
> Dear Gareth,
>
> The spm_eeg_dipole_rv.m was written by Dr. Vladimir Litvak. He provided the script in order to replying to someone' questions about ERP dipoles. You can find it by searching "dipole fit question" in the SPM mailing list. For your convenience, I attached the script.
> Compared with spm_eeg_dipole_rv.m, spm_eeg_inv_vbecd_gui.m always gives a higher variance explained. I therefore also prefer to the latter one, but I do not know whether it is the optimal one.
>
> Bests,
> Sun Delin
> ________________________________________
> From: Barnes, Gareth [[log in to unmask]]
> Sent: Thursday, January 03, 2013 7:52 PM
> To: sundelin
> Subject: RE: [SPM] Variance explained for a dipole model of ERP data
>
> Hi Sun
> I can't find spm_eeg_dipole_rv.m in my spm8 release. Could this be from a previous spm version ?
> In the meantime I would stick with the variance explained in spm_eeg_inv_vbecd_gui.
> best
> gareth
>
> ________________________________________
> From: SPM (Statistical Parametric Mapping) [[log in to unmask]] on behalf of Sun Delin [[log in to unmask]]
> Sent: 03 January 2013 02:42
> To: [log in to unmask]
> Subject: [SPM] Variance explained for a dipole model of ERP data
>
> Dear SPMers,
>
> I tried to extract the "residual variance" for a dipole model of ERP data by using the spm_eeg_dipole_rv.m. However, I found that the result was incompatible with the "Percent variance explained" worked out by the spm_eeg_inv_vbecd_gui.m. For example, the optimal solution after 10 iterations by spm_eeg_inv_vbecd_gui.m gave 94.3886% as the "Percent variance explained". However, the spm_eeg_dipole_rv.m showed 0.3878 as the "residual variance" of the dipole model after inputting the corresponding dipole coordinates and moments. I have tried several times by using different numbers of dipoles on different dataset, and always found that the two results were incompatible. I do not know which one is correct. Would you please give me some advice? Many thanks!
>
> Bests,
> Sun Delin
>
>
>
