Dear Antanas,
At the moment there is no single place where all the fields of the
structure are documented and also sometimes the quantities of interest
are not actually stored in the struct but computed on the fly so the
best practical way is what I suggested in my last message. You can
also look at the code of spm_dcm_erp. It contains some comments
explaining different DCM fields, but not all. There is possibly some
documentation scattered in other relevant DCM functions.
Regarding imaging, you can put the MNI coordinates or time of interest
in the small box with 'ms or mm' written next to it and press the
'MIP' button. It's the same as for non-DCM imaging source
reconstruction. Look at the imaging source reconstruction chapter. All
the post-processing described there you can do for imaging DCM results
as well.
Best,
Vladimir
On Fri, Jun 12, 2009 at 12:57 PM, Spokas
Antanas<[log in to unmask]> wrote:
> Dear Vladimir,
>
> Thank you so much for your fast reply, it answered my questions perfectly,
> but I do somtimes find hard to see what stands for what in DCM .mat file
> with only letter indexes for the files inside, where can I get this
> information? And the other probably quite related question: after source
> imaging reconstruction the results show only the best result for specific
> time bin, but how can I get match (variance explained) probabilities for any
> peristimulus time bin? Thank you so much again, you of great help as always!
>
> Regards,
> Antanas
>
> ________________________________
> From: Vladimir Litvak <[log in to unmask]>
> To: [log in to unmask]
> Sent: Friday, 12 June, 2009 12:05:31
> Subject: Re: [SPM] max no. of sources in DCM, log-evidence result figure?
>
> Dear Antanas,
>
> I assume you are talking about DCM for ERP. Since the graphics window
> has limited area, the graphic display indeed breaks down at some point
> if you have too many sources. This is not, however, a limitation of
> the actual DCM. From the point of view of the code, it is only limited
> by your computer's number crunching abilities, but from the point of
> view of common sense there is a limit to how many sources you can put
> in a reasonable ERP model, especially with small number of channels.
> If your models are more complicated than your data can support, you
> will usually not get significant differences in model comparison or
> you will get that the simplest model is the best because all the
> others have been penalized due to their complexity.
>
> In your case, although the connectivity matrices cannot be displayed
> properly, they are all in the DCM structure which you can find in the
> .mat file which is saved by DCM and also in the workspace once DCM
> finishes running. A good way to decipher the meaning of the different
> fields of the structure is by putting a breakpoint in
> spm_dcm_erp_results and getting to the part of the code which
> generates the figure with the thing you are interested in. By
> examining this code you can figure out how that thing is computed from
> what is stored in the DCM struct and do the same in your code.
>
> The free energy (which approximates the log evidence), for instance,
> is stored in DCM.F
>
> Best,
>
> Vladimir
>
>
>
>
>
> On Fri, Jun 12, 2009 at 10:45 AM, Spokas
> Antanas<[log in to unmask]> wrote:
>> Hey all,
>>
>> I have tried to model 5 pairs of sources in my DCM, but in GUI it looks
>> like
>> its not possible, as for instance matrix C desapears (no room it seems),
>> its
>> fine with 4 pairs (8 all together). Is there a max number in DCM and if it
>> is what is literature reference, please?
>> I have also failed to find a model log evidence number for specific model,
>> as when using BMS batch, it only gives a relative comparison, where can I
>> get the actual figure?
>> Thank you so much!!!!!!
>>
>> Regards,
>> Antanas
>>
>>
>
>
|