Dear Tom:
This contrast is, indeed, not estimable. Could it be the case that your
regressors are convolved with the HRF, or otherwise modified? Even a slight
modification would spoil the collinearity between the columns of your design
matrix. This would make your contrast estimable (but probably unreliable).
I hope this helps
/Alexandre
---------------------------------------ooooOOOO
Alexandre Andrade
Instituto de Biofísica e Engenharia Biomédica
Departamento de Física
Faculdade de Ciências da Universidade de Lisboa
Campo Grande
1749-016 LISBOA
PORTUGAL
Telefone:(351) 21 750 03 16, (351) 21 750 01 77
ext. FCUL: 20506
Fax: (351) 21 750 00 30
email: [log in to unmask]
www: http://ibeb.fc.ul.pt
OOOOoooo---------------------------------------
----- Original Message -----
From: "Tom Verguts" <[log in to unmask]>
To: <[log in to unmask]>
Sent: Monday, June 27, 2005 11:42 AM
Subject: [SPM] estimable contrasts
> Dear SPM community,
> I have a question related to estimability of contrasts. To simplify my
> design somewhat, this design consists of two runs, with two conditions
> (say C1 and C2) occurring in each run. Besides a condition-specific
> regressor in each run, I also include run-specific predictors.
> Hence, my design matrix (in simplified form) is
>
> 1 0 0 0 1 0
> 0 1 0 0 1 0
> 0 0 1 0 0 1
> 0 0 0 1 0 1
>
> First column is for C1 in run 1, second column for C2 in run 1, third
> column for C1 in run 2, fourth for C2 in run 2, and columns 5 and 6 are
> run-specific regressors.
> Now, if I ask SPM to compute a contrast for C1 in run 1 vs C1 in run 2 (a
> learning effect, contrast vector (-1 0 1 0 0 0)), (s)he calculates and
> reports the value of this contrast. However, my understanding is that this
> contrast is not estimable since it is not in the row space of the design
> matrix. But I thought SPM refuses to report contrasts that are not
> estimable? Can anyone help with this problem?
> Thanks in advance!
> Tom
|