Gui Xue -
If you are asking "what are the contrast images?", then
they are simply images of the parameter estimates ("betas")
multiplied by the contrast (have a look at them with Display).
They are numbered by the order of contrasts you entered
into your Contrast Manager.
The image for con_0002.img, created by the contrast [1 0 0 0],
is identical to the first beta*img in your case.
The image for con_0004.img, created by the contrast [1 1 0 0],
is the sum of the first two betas.
If you are asking "what do the contrasts themselves mean?",
then you need to specify whether you are testing a T-test or
F-test.
As a T-test, your first contrast [1 0 0 0] tests whether the
first beta is significantly positive. Your third contrast, [1 1 0 0],
can be thought of as testing whether the "average" of the first
two betas is significantly positive. It can be significant even if only
one or neither of the individual contrasts [1 0 0 0] and [0 1 0 0]
is significant - for example, the first beta may be zero, but the
second beta large and positive, such that their average is
still significantly bigger than zero. This would be less likely in
a conjunction, which would require both component contrasts
to be significant at a certain level (see SPM archive for more
discussion of conjunctions). Similar logic can be extended to
your contrasts 4-6.
If the contrasts entering the conjunction are orthogonal, you
are correct that no further con*imgs are created. If you want
to perform a conjunction at the 2nd-level, you need to enter
the contrasts separately into a ANOVA-type design (assuming
sphericity - again, see SPM archive).
Note that, to do a 2nd-level analysis, you need N>1 replications
of each contrast (eg N subjects, or N sessions). For a "random
effects" analysis, N typically refers to subjects. N is typically 10-20;
N=2 (the two sessions in your case) will have only 1 df and
virtually no power.
For example, you need N contrasts of the type [1 0 0 0], one
for each subject. You then do a separate SPM analysis (eg
one-sample t-test) for each such contrast. It is not meaningful
to do 2nd-level analyses over different 1st-level contrasts.
Best wishes
Rik
xuegui wrote:
> Dear Dr. Henson,
>
> Sorry to bother you once again by questiones relevant to SPM.
>
> Suppose there are two session of fMRI scan, and each have two task named as a1, a2, b1, b2 respective. I've set the following contrasts:
>
> (1) 1 0 0 0 (a1>rest) con_0002.img
> (2) 0 1 0 0 (a2>rest) con_0003.img
> (3) 1 1 0 0 (a) con_0004.img
> (4) 1 0 -1 0 (a1>b1) con_0005.img
> (5) 0 1 0 -1 (a2>b2) con_0006.img
> (6) 1 1 -1 -1 (a>b) con_0007.img
>
> and my questiones are:
>
> First, what's the meaning of con_0004.img, and also what's the relationship between con_0002.img, con_0003.img and con_0004.img.
>
> Second, it's said that we can do conjunction analysis by select contrast 1, 2 at the same time. Is it right? If it's right, as no con_000*.img were created, then how can I do the random effect analysis in the second level? And also, What the difference between the results acquired in this way and that acquired by contrast 3?
>
> Third, just very similar to the first and second question, but only change the contrast to 4, 5 and 6. As for contrast 6, it's considered as a contrast representing the main effect of factor A.
>
> I've searched these informations in the spm mail list for many time, but still not very sure about these question. Would you please be so kind to help me out? Thank you very much!
>
> Best Regards
>
> Gui Xue
> กกกกกก2002-06-24
>
> -------------------------------------------------------
> Gui Xue, Ph.D. Candidate
> Institute of Developmental Psychology
> School of Psychology
> Beijing Normal University
> Beijing, China. 100875
>
> Email: [log in to unmask]
> Phone: 86-10-62207743(O)
> Fax: 86-10-62206154
> _______________________________________________________
--
---------------------------------------------------------
DR RICHARD HENSON
Institute of Cognitive Neuroscience
& Wellcome Department of Imaging Neuroscience
University College London
17 Queen Square
London, WC1N 3AR
England
EMAIL: [log in to unmask]
URL: http://www.fil.ion.ucl.ac.uk/~rhenson
TEL1 +44 (0)20 7679 1131
TEL2 +44 (0)20 7833 7472
FAX +44 (0)20 7813 1420
MOB +44 (0)794 1377 345
---------------------------------------------------------
--
|