Li Bei wrote:
> Dear Penny:
>
> It is correct to calculate the covariance matrix
> of these regions to infer the intrisic connection
> between them?
>
No. The intrinsic connections, A, are not equal to the
covariance matrix.
They are connections in a dynamical model
of neuronal activity
dz/dt = Az
Have a look at the DCM paper for further details:
http://www.fil.ion.ucl.ac.uk/~wpenny/publications/dcm.pdf
> I can't find where is the "contrasts of
> connections" (i'm begginer,sorry)
>
>
When you 'review' a model, type in the threshold to be used
(for posterior inference) then you should prompted with
'display ...'. Select 'Contrast of connections'.
Best,
Will.
>
>
> --- Will Penny <[log in to unmask]> 5DU}ND#:
>
>>Dear Li Bei,
>>
>>1. Connectivities between regions are inferred on
>>the basis of the values of the intrinsic connections
>>(stored in DCM.A) and our uncertainty in these
>>estimates.
>>These define the posterior distribution ie. our
>>belief in the
>>values of these parameters after having fitted the
>>data.
>>Using the 'contrasts of connections' option on them
>>we can then see
>>how probable it is that a certain connection is eg.
>>greater than 0.
>>
>>Your connection A(2,1) - which I think is from
>>region
>>1 to region 2 - seems large. You can get the
>>posterior probability
>>of it being bigger than 0 as described above.
>>
>>Similarly for A(3,1).
>>
>>It may be that your strong connections are only the
>>forward ones.
>>
>>Re questions 2 and 3. These are really cognitive
>>neuroscience
>>questions which require an expertise in the
>>particular system you are studying. But generally,
>>driving
>>inputs describe stimulus bound attributes that enter
>>sensory regions
>>and modulatory connections are of a more
>>psychological nature.
>>
>>Best,
>>
>>Will.
>>
>>Li Bei wrote:
>>
>>
>>>Dear Penny:
>>> I got some other problems in DCM.
>>> While i got a set of brain regions with
>>
>>T-contrast
>>
>>>,How can i get those intrisic connections between
>>>there regions??? Since i got 3 region active
>>
>>region
>>
>>>R1,R2 and R3,i connect R1 to R2,R3 and R2 to R1,R3
>>
>>and
>>
>>>R3 to R1,R2, and put the driving input to an
>>
>>region
>>
>>>R1(i guess this might be the right input
>>
>>region,but
>>
>>>how can i validate this?). And i want to infer
>>
>>their
>>
>>>intrisic couplings by coupling coefficents between
>>>them, here is result in DCM.A
>>> -1.0000 -0.0074 -0.0810 0.0132
>>> 0.1593 -1.0000 -0.0383 -0.0017
>>> 0.4692 0.0728 -1.0000 0.0190
>>> 0.0539 -0.0095 -0.0453 -1.0000
>>>
>>>but i foud nothing in this matrix,
>>>
>>>could you help me out those question??
>>>1,how to infer connectivities between differe
>>
>>regions
>>
>>>2,which region should driving input connect to?
>>>3,which coupling should modulatory input connect
>>
>>to?
>>
>>>
>>>Thank you very much !
>>>
>>>
>>>BeiLi
>>> --- Will Penny <[log in to unmask]>
>>
>>5DU}ND#:
>>
>>>>Dear Li Bei,
>>>>
>>>>RE DCM analysis:
>>>>
>>>>When looking at the results of an SPM analysis
>>>>selecting different contrasts allows you to look
>>
>>at
>>
>>>>maps of different effects.
>>>>
>>>>In the data on the web, it was an arbitrary choice
>>>>to
>>>>look at the effect of motion. You could use other
>>>>contrasts
>>>>to look at other effects.
>>>>
>>>>Looking at 'results', however, does not cause any
>>>>VOI
>>>>files to be written. The use of contrasts in
>>>>extracting
>>>>VOIs is a separate issue.
>>>>
>>>>Here, it is useful to specify a contrast so that
>>>>effects of no interest can be removed. Think of
>>
>>this
>>
>>>>as a filtering. So, if I remember correctly, an
>>>>F-contrast over the effects of Photic, Motion
>>>>and attention was used as we are interested in
>>>>modelling *all* of these effects with DCM.
>>>>
>>>>If we were only interested in modelling a subset
>>
>>of
>>
>>>>these
>>>>effects we would have used an F-contrast spanning
>>>>only
>>>>that subset.
>>>>
>>>>RE GLM analysis:
>>>>
>>>>Perhaps you could simplify your design by having
>>>>only one
>>>>'control' variable (for A and B), instead of 3
>>>>(control for A, control for B,
>>>>control for A and B). I guess its possible that [1
>>>>-1 0 0 0 0]
>>>>could also be formed by a combination of the other
>>>>regressors (which
>>>>would result in an invalid contrast) - but
>>>>I have'nt worked this out in detail.
>>>>
>>>>Best wishes,
>>>>
>>>>Will.
>>>>
>>>>Li Bei wrote:
>>>>
>>>>
>>>>
>>>>>Dear Penny:
>>>>> Thank you for you help,you are so kind!
>>>>> Sorry for bringing you so much troubles.
>>>>> I have another 2 quesion :)
>>>>>QUESTION 1
>>>>>BEGIN
>>>>> What's the difference of VOI in chosing
>>>>
>>>>different
>>>>
>>>>
>>>>>F-contrast in DCM instruction published on you
>>>>>websit,the Result section.
>>>>> In you example,you chosed Motion Condition
>>>>
>>>>,and
>>>>
>>>>
>>>>>defined VOIs,but i found,value series of VOI are
>>>>>different in different
>>>>>F-contrast(Photic,Motion,Attention),why you chose
>>>>>Motion in you example?? What will happen if you
>>>>
>>>>chose
>>>>
>>>>
>>>>>Photic and define VOIs?
>>>>>END
>>>>>
>>>>>
>>>>>
>>>>>QUESTION 2
>>>>>BEGIN
>>>>>This Question is about fMRI modelling :)
>>>>>
>>>>>My exp. is designed like this
>>>>>CACBCACBCACB (12 total) and A for stimuli A,B
>>>>
>>>>for
>>>>
>>>>
>>>>> stimuli B,C for control
>>>>>there is 10 scans for each character ,so there
>>
>>are
>>
>>>>120
>>>>
>>>>
>>>>>scans totally
>>>>>
>>>>>in fMRI Design Stage,i set the paramaters like
>>>>
>>>>this#:
>>>>
>>>>
>>>>>scans per session[120]
>>>>>Number of conditions[6]
>>>>>
>>>>>name for condition 1 [StimuliA]
>>>>>Vector [10 50 90]
>>>>>
>>>>>name for condition 2 [StimuliAControl] //Control
>>>>>Before Stimuli A
>>>>>vector [0 40 80]
>>>>>
>>>>>name for condition 3 [StimuliB]
>>>>>vector [30 70 110]
>>>>>
>>>>>name for condition 4 [StimuliBControl] //Control
>>>>>Before Stimuli B
>>
> === message truncated ===
>
> =====
> Bei Li
> Institute of Neuroinformatics
> Dalian University of Technology
> 2 Ling Gong Rd, Dalian 116023, China
> Email:[log in to unmask]
>
> _________________________________________________________
> Do You Yahoo!?
> W"2aJ@=gR;AwF7VJ5DQE;"Cb7Q5gSJ
> http://cn.rd.yahoo.com/mail_cn/tag/1g/*http://cn.mail.yahoo.com/
>
>
--
William D. Penny
Wellcome Department of Imaging Neuroscience
University College London
12 Queen Square
London WC1N 3BG
Tel: 020 7833 7475
FAX: 020 7813 1420
Email: [log in to unmask]
URL: http://www.fil.ion.ucl.ac.uk/~wpenny/
|