Dear brain mappers
I am running a 2nd level analysis using a one-way ANOVA. I have four
groups. I have made non-spericity correction, since the groups are four
different effects measured in the same 14 subjects.
If I make the t-contrast [1 -1 0 0] I assume the estimate of the first
group will we larger than for the second group in the significant voxels in
my SPM{T} map. I now choose a voxel (in this case the global max of the
contrast [1 -1 0 0]) and make a plot of the contrast estimates. I
choose 'effects of interest' (i.e. the one generated by default after the
estimation (I can see this is not a diagonal matrix))) in order to get a
display of all the four estimates. To my surprise it seems that the
estimate of the second group is LARGER than the estimate for the first
group (even though the voxel is picked from a thresholded map of the [1 -1
0 0] contrast). This is displayed on page 1 of the attached document. I
assume this has something to do with the non-sphericity correction.
I now create an 'effects of interest' as a diagonal matrix (F-contrast) [1
0 0 0; 0 1 0 0; 0 0 1 0; 0 0 0 1].
If I then make a plot of the contrast estimates and use the diagonal
matrix, the contrast estimates now look more sensible (group 1 is larger
than group 2 as indicated by the contrast I'am testing [1 -1 0 0]). See
page 2 of the attached document
My questions:
If I want to report the contrast estimates for all four groups in a
particular voxel, can I then use the default 'effects of interest' created
by SPM or should I use the diagonal matrix I have created my self in order
to report something that is physiological sensible, i.e. "In group 1 there
is a larger response in voxel x than in group 2"
If I can use default 'effects of interest', how should I then interpret the
sizes of the estimates?
If I use my 'own' diagonal 'effects of interest', how should I then
interpret the sizes of the estimates?
I hope some of the experts could clarify this point and maybe explain the
default 'effects of interest' when non-spericity correction is used.
Sincerely,
Mark S. Christensen
MSc, PhD-student
Danish Research Centre for Magnetic Resonance
Copenhagen University Hospital Hvidovre
&
Institute of Physical Exercise and Sport Science
University of Copenhagen
Denmark
|