You can actually just compute one one sample t-test. Just create 2 contrasts (e.g. 1 and -1) for the second level. This will give you the same results as Julie suggests, but with less work.
It depends on the conditions that were defined at the first level for each subject. Let's assume they are Rest, Simple and Complex. You would define at least the contrasts for:
1. Complex minus Simple:
0 -1 1
2. Simple minus Complex:
0 1 -1
Then do a one-sample t-test using *.con files form contrast 1 to get activations greater in Complex and another one-sample t-test using *.con files from Contrast 2 for activations greater in Simple condition.
Hope this helps-
Julie
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
From: SPM (Statistical Parametric Mapping) [mailto:[log in to unmask]] On Behalf Of ???
Sent: Sunday, June 29, 2008 12:07 AM
To: [log in to unmask]
Subject: [SPM] Questions regading fMRI analysis
Dear experts,
Sorry for bothering you with basic questions.
I'm conducting an fMRI study of simple and complex motor activities in a group of patients using a block design.
I hypothesized that some brain regions may show more activation during complex motor activity as compared to simple motor activity in the same subjects (N = 13).
And here is my first question, "which statistical model should I select in the second-level analysis of simple versus complex motor activity ?, two-sample t-test or paired t-test ?"
My preliminary results showed that in the second-level analysis using one-sample t-test, expected brain regions (e.g., M1, SMA, cerebellum) were strongly activated at FWE-corrected P < 0.05 both in simple and complex motor activities.
However, second-level analysis using paired t-test (simple vs. complex activity) revealed that some clusters have only survived uncorrected P < 0.001, which were not significant at FWE or FDR-corrected P value.
I am wondering whether or not this result of difference between simple and complex activities is meaningful.
Many thanks for your kind instructions.
Best wishes,
Kim