Hi Gang. I've pasted my 3dANOVA2 script below (and my 3dttest script for comparison), but I'm pretty sure it shouldn't be the same as a paired t-test. In a paired t-test, for each subject, I'm calculating the difference between (presumably) scans 1 and 3 (the first pre- and first post-intervention scans), and between scans 2 and 4 (the second). That's how I've set up the paired t-test, anyway, and the resulting dataset has 9 degrees of freedom (because I have 5 subjects, each of whom has two paired scans, for a total of 10 pairs). It's a little weird as a comparison because it doesn't treat the four scans as though they're all from the same subject. In the ANOVA, I can treat all four scans as observations from the same subject, two in each condition (pre- and post-intervention), which was the attraction in the first place. In that case, according to the ANOVA manual, the ttest from -adiff has degrees of freedom equal to the number of b levels, i.e. the number of subjects, i.e. 5, minus 1. This is why 3dMEMA seemed like it might be particularly appropriate -- I'm not sure how I'd set up this particular situation in a t-test.
So what I'm asking about reportability is less how the statistic is calculated, which I know, and more whether there's a widely recognized way to identify it. I can certainly reproduce the formula if it's down to that; I just fear that reviewers will get worried needlessly.
Anyway, hope that clarifies things. Thanks for your response!
===============
3DANOVA2 SCRIPT
#!/bin/sh
sm=$1
cd /jet/mweb/maxwell/data/aggregate
3dANOVA2 \
-DAFNI_FLOATIZE=YES \
-type 3 \
-alevels 2 \
-blevels 5 \
-dset 1 1 PCASL4.blur${sm}.CBF0.at.s1.mean+tlrc \
-dset 1 2 PCASL4.blur${sm}.CBF0.at.s2.mean+tlrc \
-dset 1 3 PCASL4.blur${sm}.CBF0.at.s3.mean+tlrc \
-dset 1 4 PCASL4.blur${sm}.CBF0.at.s4.mean+tlrc \
-dset 1 5 PCASL4.blur${sm}.CBF0.at.s8.mean+tlrc \
-dset 1 1 PCASL3.blur${sm}.CBF0.at.s1.mean+tlrc \
-dset 1 2 PCASL3.blur${sm}.CBF0.at.s2.mean+tlrc \
-dset 1 3 PCASL3.blur${sm}.CBF0.at.s3.mean+tlrc \
-dset 1 4 PCASL3.blur${sm}.CBF0.at.s4.mean+tlrc \
-dset 1 5 PCASL3.blur${sm}.CBF0.at.s8.mean+tlrc \
-dset 2 1 PCASL2.blur${sm}.CBF0.at.s1.mean+tlrc \
-dset 2 2 PCASL2.blur${sm}.CBF0.at.s2.mean+tlrc \
-dset 2 3 PCASL2.blur${sm}.CBF0.at.s3.mean+tlrc \
-dset 2 4 PCASL2.blur${sm}.CBF0.at.s4.mean+tlrc \
-dset 2 5 PCASL2.blur${sm}.CBF0.at.s8.mean+tlrc \
-dset 2 1 PCASL1.blur${sm}.CBF0.at.s1.mean+tlrc \
-dset 2 2 PCASL1.blur${sm}.CBF0.at.s2.mean+tlrc \
-dset 2 3 PCASL1.blur${sm}.CBF0.at.s3.mean+tlrc \
-dset 2 4 PCASL1.blur${sm}.CBF0.at.s4.mean+tlrc \
-dset 2 5 PCASL1.blur${sm}.CBF0.at.s8.mean+tlrc \
-mask intracranial+tlrc \
-fa pre-post-stim \
-fab stim-subj-ixn \
-acontr 1 -1 stim-contrast-post-vs-pre \
-adiff 1 2 stim-diff-post-vs-pre \
-bucket CBFANOVA2_stim
cd /jet/mweb/maxwell/scripts
===============
3DTTEST SCRIPT
#!/bin/sh
cd /jet/mweb/maxwell/data/aggregate
sm=$1
3dttest \
-prefix PCASL.blur${sm}.CBF0.t.stim \
-paired \
-set1 \
PCASL1.blur${sm}.CBF0.at.s1.mean+tlrc \
PCASL2.blur${sm}.CBF0.at.s1.mean+tlrc \
PCASL1.blur${sm}.CBF0.at.s2.mean+tlrc \
PCASL2.blur${sm}.CBF0.at.s2.mean+tlrc \
PCASL1.blur${sm}.CBF0.at.s3.mean+tlrc \
PCASL2.blur${sm}.CBF0.at.s3.mean+tlrc \
PCASL1.blur${sm}.CBF0.at.s4.mean+tlrc \
PCASL2.blur${sm}.CBF0.at.s4.mean+tlrc \
PCASL1.blur${sm}.CBF0.at.s8.mean+tlrc \
PCASL2.blur${sm}.CBF0.at.s8.mean+tlrc \
-set2 \
PCASL3.blur${sm}.CBF0.at.s1.mean+tlrc \
PCASL4.blur${sm}.CBF0.at.s1.mean+tlrc \
PCASL3.blur${sm}.CBF0.at.s2.mean+tlrc \
PCASL4.blur${sm}.CBF0.at.s2.mean+tlrc \
PCASL3.blur${sm}.CBF0.at.s3.mean+tlrc \
PCASL4.blur${sm}.CBF0.at.s3.mean+tlrc \
PCASL3.blur${sm}.CBF0.at.s4.mean+tlrc \
PCASL4.blur${sm}.CBF0.at.s4.mean+tlrc \
PCASL3.blur${sm}.CBF0.at.s8.mean+tlrc \
PCASL4.blur${sm}.CBF0.at.s8.mean+tlrc
cd /jet/mweb/maxwell/scripts
