Dear Sanja,
a 1st level F-test will NOT produce any contrast image, only an F-image
and ESS-image.
So I wonder what you passed to the 2nd level following your 1st level ?
Simply do a t-test at the 1st level, this will produce a con-image and a
t-image. Then use the con-images at the 2nd level.
As expected, 1st level contrast [1 -1] followed by a 2nd level
t-contrast [1] will give you the same results as 1st level contrast [-1
1] and 2nd t-contrast [-1].
To test "both directions" at the 2nd level, use an [1] F-test at 2nd
level with [1 -1] (or [-1 1]) t-contrast at 1st level.
HTH,
Chris
===================================================
Christophe Phillips, Ir, PhD
FRS-FNRS Research Associate
Adjunct assistant professor in applied sciences
Cyclotron Research Centre, B30
University of Liege, Sart Tilman
4000 Liege, Belgium
Tel: +32 4 366 2316 (secr.)
+32 4 366 2366
Fax: +32 4 366 2946
email: [log in to unmask]
web: http://www.cyclotron.ulg.ac.be
===================================================
Le 21/05/2014 19:17, Sanja Nedic a écrit :
> Dear SPM Experts,
>
> I have a somewhat basic question regarding T and F contrasts in SPM 8.
>
> Specifically, I have a first level design with 12 conditions, and
> would like to simply compare the first 6 to the last 6 (since the
> first 6 are obtained under one drug, and the remaining 6 under another).
>
> At the first level, I've run both a T test with contrast matrix: 1 1 1
> 1 1 1 -1 -1 -1 -1 -1 -1 (positive effect of the first drug) and an F
> test with contrast matrix of the exact same form: 1 1 1 1 1 1 -1 -1 -1
> -1 -1 -1 (main effect of the drug). At the second level, I am just
> running a one-sample T test (t test with contrast matrix 1 or -1) on
> con/ess images coming from the first level (one image per subject).
>
> I always had the impression that if I run both directions of the T
> test, either at the first, or at the second level (that is, either run
> T test with contrast matrix:1 1 1 1 1 1 -1 -1 -1 -1 -1 -1 at the 1st
> level followed first by one sample T test with contrast 1 at the 2nd
> level and then by one sample T test with contrast:-1 at the second
> level, or first run T test with contrast matrix:1 1 1 1 1 1 -1 -1 -1
> -1 -1 -1 and T test with contrast matrix: -1 -1 -1 -1 -1 -1 1 1 1 1 1
> 1 at the first level, both followed by one sample T test with contrast
> matrix: 1 at the 2nd level), the significant differences should sum up
> to significant differences found by running the F test at the first
> level and a (one sample) T test with contrast 1 at the second.
>
> However, this does not seem to be the case. The results obtained using
> two directions of the T test do not sum up to the results obtained
> using the F test. Please see the attached image. I am getting
> significant differences by using the F test, as well as by using one
> direction of the T test, but not the other. In addition, 2nd level
> significant differences found using the T test at the first level do
> not match the differences found using the F test at the first level.
>
> Could you please tell me why this would be the case?
>
> Thank you.
>
> Best,
> Sanja
>
> --
> Sanja Nedic
> Ph.D. Candidate
> Department of Biomedical Engineering
> Bioengineering Building
> Stony Brook University
> Stony Brook, NY 11794-5281
