minidragon wrote:
> Dear Will, Thank you for your help. I checked my data according to what you mentioned in the
> list. In fact I followed the first two rules. In terms of the third, T contrast is used when I
> extracted VOI data. And I also tried the F contrast, but the results weren't improved. I think
> maybe my experiment design isn't fit for DCM analysis because it isn't a real factor design. Now
> I describe the design in brief. It is a very simple block design with two conditions A and B and
> the baseline is a plus. The stimulus of A and B were the same. During condition A, the subject
> was required to judge by pressing a button. For B, the subject only need to press a button
> without judgement. And I want to observe whether judgement in A could hava a modulatory effect.
> When specifying the DCM model, the driving input included both condition A and B. That is both
> conditions were combined into one column in design matrix. And it was supposed to have an effect
> on Brodmann 17/18 area. Seldom activation was found in BA17 in my data. So the area I selected is
> BA18. The modulatory input contained only conditon A. Is there something wrong with the driving
> input?
So, echoing my earlier comment, did an F-contrast over the driving input activate
BA18 in your GLM analysis. If not, then something is 'wrong' in that a DCM analysis
won't be appropriate if the brain regions you are measuring from don't show
strong experimental effects.
> Only the driving input connection was always very slow. I still can't think it very clear.
> One question: If the correlation between the time course extracted from V1/V2 and the driving
> input is higher, the input connection will be larger?
Yes, that's right.
Best,
Will.
Thanks again. Best wishes. Yours
> minidragon. ----- Original Message ----- From: "Will Penny" <[log in to unmask]> To:
> "minidragon" <[log in to unmask]> Cc: <[log in to unmask]> Sent: Tuesday, February 01, 2005 7:38
> PM Subject: Re: a question about DCM
>
>
>
>> Dear Minidragon,
>>
>> Here's a list of things to check:
>>
>> 1. Are you using the same variable as a direct input and a modulatory input ?
>>
>> If so, this is not a good idea as its effect cannot be disambiguated in this way.
>>
>> 2. Does the variable you've specified as a direct input show a significant effect in a GLM
>> analysis ?
>>
>> If not, its unlikely you'll see an effect with DCM.
>>
>> DCM is really just a different way of explaining the same effect. If there's no effect there to
>> explain there's nothing DCM can do.
>>
>> 3. What contrast did you use when extracting the VOI data ? Its important that you use a
>> contrast that preserves the effect you are interested in but removes effects of no-interest.
>> Usually, an F-contrast over the effects of interest is appropriate.
>>
>> Best wishes,
>>
>> Will.
>>
>> minidragon wrote:
>>
>>
>>> Dear Will: I met a problem when I analysis my fMRI data of block design with DCM. I found
>>> that no matter what model I specified, the posterior probability of the input connection is
>>> very small(<0.02) and sometimes even equal to zero.That means that the input doesn't have any
>>> effect to the v1/v2 region. And the posterior probability of the other connections such as
>>> modulatory and intrinsic connections are close to one. I feel puzzled why the input
>>> connection is always so small. The regions or the driving input were wrong? The experiment
>>> isn't fit for DCM analysis, Or the models selected aren't good? By the way,my experiment
>>> design is similar to what is described in
>>> http://www.jiscmail.ac.uk/cgi-bin/wa.exe?A2=ind04&L=spm&P=R237160&I=-1
>>> <http://www.jiscmail.ac.uk/cgi-bin/wa.exe?A2=ind04&L=spm&P=R237160&I=-1>
>>> <http://www.jiscmail.ac.uk/cgi-bin/wa.exe?A2=ind04&L=spm&P=R237524&I=-1> by Darren. The
>>> baseline of the experiment was viewing a plus in the center of a screen. Thank you in
>>> advance. Good luck. Yours Minidragon
>>
>> -- 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/
>>
>>
--
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/
|