> Here is my understanding on the conjunction in second level analysis. I first establish a pair-t test model in second level, and select the con_0002.img and con_0003.img for each subject (suppose N>10). Then I estimate this model and set new contrast, say 1 0, and 0 1. And then by doing conjunction analysis, what I should do is just select the two contrasts together. Am I right?
You won't be able to do [1 0] or [0 1] contrast in a paired t-test,
because the design matrix is rank deficient. You will need to create
a design matrix without a constant term. I can't remember whether
this option is in SPM99 release - but if not, check the SPM archive
for how to do this.
> There is one more question relevant to this issue. That is, some SPMer in the list said that the contrast 1 0 and 0 1 is not orthogonal, it's not acceptable in the second level(right?). I remember some codes also posted in the archive. But I'm now not very sure how to do it. Say, where can I find the correct code and how can I add in my spm program. Thank you very much!
The contrast VECTORS are orthogonal, but you are correct that the
contrasts-multiplied-by-design-matrix (X*c) may not be orthogonal
(eg if the regressors are correlated) - which is what is required for
conjunctions. You will know this simply by whether SPM creates a
new "orthogonalised" contrast when you request a conjunction. If it
does, then take the orthogonalised con*imgs to the 2nd-level (eg,
if you conjoin [1 0] and [0 1], and SPM creates a third contrast, eg
[-0.2 0.8], then you could take con_0001.img and con_0003.img
to the 2nd-level).
> Another question is I'm now do a ROI base intensity analysis. one way to calculate the intensity is using the spmT_000* file, and the other is using con_000* file. The first file contains T value for each voxel, while the last file contains beta file. And my question is which one is better to represent the "Intensity".
"Intensity" is represented in the con*imgs, not the spm*imgs.
The spm*imgs are statistics, not parameter estimates (ie they
are the ratio of a contrast of parameter estimates over the
standard error of that contrast). Note however that the units
of the values in the con*imgs depends on the scaling of the
covariates - see SPM archive about "% signal change".
Rik
--
---------------------------------------------------------
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
---------------------------------------------------------
--
|