Hi, good point, but no, this is fine. In fact, when you consider more
fully what's going on, e.g.:
A1 = a + b + m1
A2 = a + b + m2
B1 = -a + m1
B2 = -a + m2
C1 = -b + m1
C2 = -b + m2
A = (A1 + A2) / 2 = a + b + (m1 + m2)/2
B = (B1 + B2) / 2 = -a + (m1 + m2)/2
C = (C1 + C2) / 2 = -b + (m1 + m2)/2
(A+C)/2 - B = a/2 + a = 3a/2
whereas just B also includes the subject means (m1, m2) so isn't the
same. Hence you do want the [1.5 0 0 0 ...] I think.
If you don't get any significant activation, you should check that the
unthresholded z-stat does at least look sensible (including its
histogram) - just in case I'm wrong about this ;-)
Cheers.
On 11 Apr 2008, at 13:53, Lipperts Matthijs (CTC) wrote:
> Hi!
> Thanks for the response.
> This was also my first idea, but then the contrast (A+C)/2-B would
> be the same as -1.5 B, which is strange to say the least. Also when
> I ran FSL I didn't get any significant activation.
> My concern is thus also if ány addition can be done with the two
> vectors (EV1 EV2). To compare with the normal case of 2 time points
> (A,B): in that case only 1 vector is used (1, -1), with contrasts
> [1,0,0,0....] or [-1, 0 0 0 0 ] to form either A-B or B-A. A+B
> cannot be made this way (please correct me if I am wrong). The same
> way A+C cannot be formed in the triple-case?
> One (maybe too simple) method I thought of was to use the vector [1
> -2 1] (instead of the standard triple t-test [1 -1 0] and [1 0 -1]),
> and contrast (1,0,0,0,0..). This way i found some activation, but
> before interpreting it I would like to be sure that this method is
> valid.
> Thanks again,
> Matthijs
>
>
> -----Original Message-----
> From: FSL - FMRIB's Software Library on behalf of Steve Smith
> Sent: Fri 4/11/2008 9:21 AM
> To: [log in to unmask]
> Subject: Re: [FSL] setting up contrast to incorporate time effects
>
> Hi, from the triple-t-test example in the FEAT manual:
>
> A=a+b
> B=-a
> C=-b
>
> Hence if you want
>
> (A+C)/2-B
>
> you get
>
> (a+b-b)/2+a = 3a/2
>
> so you want the contrast [1.5 0 0 0 0.....]
>
> Cheers.
>
>
> On 10 Apr 2008, at 09:15, Matthijs Lipperts wrote:
>> Dear all,
>>
>> We are currently working on a project to see the effects of a drug
>> (botulinum toxin) on the activation of some brain area's. However to
>> prove
>> that the decrease in signal is not a time-effect or training-effect,
>> the
>> measurement is repeated 7 weeks after the injection. We thus have 3
>> time
>> points: A-->before injection, B--> after injection, C--> 7 weeks
>> after
>> injection.
>> To see group results I followed the 'tripled T-test' instructions
>> with the
>> contrasts for A-B, B-C and A-C, and this works perfectly.
>> However, we would very much like to see the contrast (A+C)/2-B,
>> since this
>> gives us the difference between the time effect and the drug-effect.
>> So far
>> I have not succeeded in this. Any suggestions are helpful!
>> Thank you in advance,
>> Sincerely, Matthijs
>>
>
>
> ---------------------------------------------------------------------------
> Stephen M. Smith, Professor of Biomedical Engineering
> Associate Director, Oxford University FMRIB Centre
>
> FMRIB, JR Hospital, Headington, Oxford OX3 9DU, UK
> +44 (0) 1865 222726 (fax 222717)
> [log in to unmask] http://www.fmrib.ox.ac.uk/~steve
> ---------------------------------------------------------------------------
>
---------------------------------------------------------------------------
Stephen M. Smith, Professor of Biomedical Engineering
Associate Director, Oxford University FMRIB Centre
FMRIB, JR Hospital, Headington, Oxford OX3 9DU, UK
+44 (0) 1865 222726 (fax 222717)
[log in to unmask] http://www.fmrib.ox.ac.uk/~steve
---------------------------------------------------------------------------
|