For second-level (group) analysis, the contrast weights should sum to zero.
For example, contrast weights for group 1 > group 2 would be 1 -1 (the
reverse for group 1 < group 2).
As to the "third" column- did you include a covariate? If the covariate is
of no interest, it should be given a weight of zero (0). The contrast
weights for group 1 > group 2 would be 1 -1 0, and so on.
Julie E. McEntee, M.A., C.C.R.P.
Senior Research Program Coordinator
Department of Psychiatry and Behavioral Sciences
Division of Psychiatric Neuroimaging
Johns Hopkins University School of Medicine
600 N. Wolfe St./ Phipps 300 (office: room 317)
Baltimore, MD 21287
Phone: 410-502-0468
Fax: 410-614-3676
-----Original Message-----
From: SPM (Statistical Parametric Mapping) [mailto:[log in to unmask]] On
Behalf Of So-yeon Kim
Sent: Tuesday, August 19, 2008 1:03 PM
To: [log in to unmask]
Subject: [SPM] two sample t-test and contrast
Hi SPMers,
I have a question about defining contrast with 2 sample t-test.
I've searched the e-mail list for this topic, but I still have questions
about how the contrast works in the two sample t-test.
I have an event-related design with 2 different groups. Two groups performed
the same tasks with same conditions. In my second level analysis, I chose
the two-way sample t-test to compare group differences between 2 groups on
the same experimental condition. After I estimated my SPM.mat, I defined
contrast for group 1 > group 2, group1 < group2. When I was making the
contrast, however, I had a question about the last column of my contrast
map. I have 3 columns instead of 2. First, I thought I could put 'zero' for
this column, but when I tried to see the average activity for 2 groups, by
giving a value 1 for 'group 1' and giving a value 1 for 'group 2', I noticed
that I had to input '2', instead of 'zero' for the last column. Does the
last column represent between group effects? It seems that I should insert
the sum of the contrast value in the last column, to make the contrast work
(e.g. "1 0 1", "1 -1 0", "1 1 2").
Any answers will be deeply appreciated!!
Also, I compared the results for activation in each group (from the separate
analysis of one-way t-test for each group) to those from this 2 sample
t-test (by giving contrast "1 0 1" or "0 1 1". The results from those
different estimation looked similar in terms of intensity and location of
the activation, but they are not exactly identical. It was true for the
average activation as well. Which one should I use for reporting the
results?
I appreciate your knowledge and help in advance and good luck with your
research!
Best,
So-yeon
|