Steve,
sounds like you want the interaction between time and group.
I'd suggest setting up a factorial anova. you have a within subjecs factor--time and a between subjects factor--group. choose anova no constant.
In SPM2 you will need to specify non-sphericity over the within factor. (SPM5 is set up nicely for these kinds of analyses with between and within factors)
assuming you enter groups in this order:
A1 A2 B1 B2
1 -1 -1 1 is the contrast you are interested in. this is the interaction. an F-test will tell you where groups differ with respect to change over time.
-1 1 -1 1 as a t-contrast will tell you the main effect of time (where activation increases over time) regardless of group.
1 1 -1 -1 and
-1 -1 1 1 are main effects of group.
its also useful to manually specify an unweighted effects of interest F contrast before you look at your interaction.
1
0 1
0 0 1
0 0 0 1
This way, if any significant blobs show up in the interaction contrast you can plot them easily and see which way the interaction is occuring.
Also, when looking at the main effects (if those are of interest to you) you should exclusive mask any significant interaction blobs. (in results when SPM asks if you want to 'mask with other contrasts' say yes and choose the interaction.
I hope this helps
scj
Sterling C. Johnson, Ph.D.
Assistant Professor of Medicine
University of Wisconsin Medical School
Madison VA Hospital
GRECC
2500 Overlook Terrace (11G),
Madison WI 53705
phone 608-256-1901 x 11946
fax 608-280-7165
email: [log in to unmask]
www.brainmap.wisc.edu
>>> "Cramer, Steven" <[log in to unmask]> 02/04/07 3:54 PM >>>
May I ask help with operational details in SPM2?
Two groups of subjects were each studied twice over time, and the
fMRI data were analyzed with standard RFX methods. We are able to
examine the change over time in each group, with either paired
t-testing or Imcalc [i2-i1] approaches.
The goal is to determine whether the change in one group is different
from the change in a second group. The crux of the Q is that change
is examined in a 2-tailed way, could be up or down.
We approached the goal by comparing the [i2-i1 maps], by group, using
a 2-sample t-test. However, significant activation present in maps
generated with this approach could indicate that the first group had
relatively greater increase over time, or it could indicate that the
second group had relatively greater decrease over time.
What steps would one take in SPM2 to disentangle these two
possibilities? Thanks,
Steve
|