Hi Chris and Vladimir,
with DCM/MEEG, model comparison of models with different number of
sources is possible.
However, you have to use a script, because this is not yet implemented
in the GUI. An example script, which can be extended can be found in the
man/example_scripts in the SPM8 distribution.
With scripts, you can use this recipe:
Identify a model that includes all the sources of the other models.
Specify this model first. In the DCM struct, there will be matrix
DCM.M.E (called E_max in the following) which contains the spatial modes
of this model. If you now specify other models with less sources, you
need to replace after model specification, and before model inversion,
DCM.M.E by E_max. DCM will then use the same set of spatial modes for
both models.
We used this technique in the Fastenrath, Friston, Kiebel, NeuroImage
2009 Jan 1;44(1):154-63, 'Dynamic causal modelling for M/EEG: spatial
and temporal symmetry constraints.' In this paper, we compared 4-source
with 2-source models.
All the best, Stefan
> Hi Chris,
>
> On Fri, Nov 20, 2009 at 9:44 AM, Christophe Phillips
> <[log in to unmask]> wrote:
>
>> I see the point about using the leadfield for data reduction BUT this
>> hampers one key (at least to me) feature of DCM-EEG: the possibility to
>> compare models with different source configurations (number/location)!
>>
>
>
> That's a very good point that I wanted to mention in my reply to Zoi,
> but didn't have time for it yesterday. With the way things are
> presently implemented to compare models with different sources one
> needs to include all those sources in all the models so that the
> models will only differ in the connectivity part. If you don't connect
> some of the sources to any inputs, these sources will be silent and
> will not be able to fit any data, but they will still count for
> reduction to modes. So for instance if you want to compare between two
> models which differ in location of a particular source, put sources in
> both alternative locations and then connect either one or the other to
> the rest of the network.
>
> When in doubt about the validity of your model comparison, you can use
> the 'Verify data identity' option in the BMS GUI. This option is now
> off by default but if you turn it on, it'll only allow valid model
> comparisons to be performed.
>
> Best,
>
> Vladimir
>
>
>
>
>> Vladimir Litvak a écrit :
>>
>> Dear Zoe,
>>
>> I briefly looked at the code and it seems that what it does now is not
>> exactly the same as what was done when those papers were being
>> prepared. Now the spatial modes are based on the leadfields of the
>> sources included in DCM models rather than on the data. So there is no
>> place in the code where the percentage of data variance explained by
>> these modes is computed. This could possibly be done with some custom
>> code but writing such code would require some expertise.
>>
>> Best,
>>
>> Vladimir
>>
>> On Thu, Nov 19, 2009 at 3:50 PM, Zoi Da <[log in to unmask]> wrote:
>>
>>
>> Dear Vladimir,
>>
>> My mistake for not explaining.
>>
>> In various articles of Garrido et al. (2007;2008;2009), they state that
>> 'the use of three principal eigenvariates (spatial modes) preserved more
>> than 73% of the variance in all subjects.' I was wondering how someone can
>> test that? I have used 8 spatial modes, how can i test how much variance
>> they explain in al subjects? Where are these values stored?
>>
>> Hope that it is clearer now!
>>
>> Best
>>
>> Zoi
>>
>> P.S. Thanks for taking the time
>>
>> ________________________________
>> Von: Vladimir Litvak <[log in to unmask]>
>> An: Zoi Da <[log in to unmask]>
>> CC: [log in to unmask]
>> Gesendet: Donnerstag, den 19. November 2009, 15:23:17 Uhr
>> Betreff: Re: [SPM] WG: DCM_EEG
>>
>> Dear Zoi,
>>
>> It's not quite clear to me what you want to do. Perhaps you could
>> describe the aim of your analysis and then it's be easier to help you.
>>
>> Best,
>>
>> Vladimir
>>
>> 2009/11/19 Zoi Da <[log in to unmask]>:
>>
>>
>> ----- Weitergeleitete Mail ----
>> Von: Zoi Da <[log in to unmask]>
>> An: [log in to unmask]
>> Gesendet: Donnerstag, den 19. November 2009, 14:56:12 Uhr
>> Betreff: DCM_EEG
>>
>> Hi everyone
>>
>> My name is Zoi and I have been analyzing EEG data with DCM. I have
>> computed
>> the DCMs to eight spatial modes. I would like know how I can compute the
>> variance of the data?
>>
>> Thanks a lot
>>
>>
>> __________________________________________________
>> Do You Yahoo!?
>> Sie sind Spam leid? Yahoo! Mail verfügt über einen herausragenden Schutz
>> gegen Massenmails.
>> http://mail.yahoo.com
>> __________________________________________________
>> Do You Yahoo!?
>> Sie sind Spam leid? Yahoo! Mail verfügt über einen herausragenden Schutz
>> gegen Massenmails.
>> http://mail.yahoo.com
>>
>>
>> __________________________________________________
>> Do You Yahoo!?
>> Sie sind Spam leid? Yahoo! Mail verfügt über einen herausragenden Schutz
>> gegen Massenmails.
>> http://mail.yahoo.com
>>
>>
>>
>>
>
>
--
Dr. Stefan Kiebel
Max Planck Institute for
Human Cognitive and Brain Sciences
Leipzig, Germany
Phone: ++49 341/9940-2435
Fax: ++49 341/9940-2221
http://www.cbs.mpg.de/~kiebel
|