Print

Print
Sounds like you probably ended up running two different models.  Instead of using the model wizard, I suggested setting up the model by hand using the tips here
http://www.fmrib.ox.ac.uk/fsl/feat5/detail.html#PairedTwoGroupDifference

Then you'll know *exactly* how the model was set up.  Here are the answers to your specific questions

1) No
2) Statistically, not a problem but interpretation might be off.
3) I don't really know.  I guess you can run a second model that models the paired difference for your 9 months apart group separate from the same day group and see if the paired differences differ between the two groups.  Sample sizes within the two groups are pretty small, so it is sort of hard to interpret any differences you see.  You're mostly hoping there isn't a difference. 

Jeanette

On Tue, Oct 12, 2010 at 12:33 PM, Estephan Moana <[log in to unmask]> wrote:
Hello all, I did a paired 2-group difference analysis and got unexpected results that I hope you can help me understand. I had a group of 12 subjects imaged under 2 levels of stimulation (non-painful pressure/painful pressure). When I ran the analysis with condition A being "painful pressure" and B "non-painful pressure", I got no activation for contrast "A > B" and several activations for contrast "A < B" - when I was expecting that "A > B" would show way more activations than "A < B". To verify these results, I re-ran the analysis now entering "non-painful pressure" as condition A. This time both contrasts had activations listed, when I expected the results to be same as the first analysis but only inverted (activations for "A > B", no activation for "A < B"). I did use the "model setup wizard - 2 groups, paired", and verified my inputs so that they would be in the correct order.

My questions are:
1) Is it possible to have different results by changing the input feat folders order of condition "A" and "B"?
2) 5 subjects had one condition imaged 9 months apart from the second condition. The remaining 7 subjects had both conditions imaged in the same day. Can this play a role in the results?
3) Any ideas how can I further explored these results, if they are not an error byproduct?

Best regards.

Estephan Moana, D.D.S., M.S.
Oral Biology PhD Program - Neurobiology track
Center for Neurosensory Disorders
School of Dentistry, UNC- Chapel Hill