Print

Print
Hi - I'm not completely sure what you're saying, so there's at least 3 possibilities that come to mind:

1) If one set of results is the average from the group, then it's not surprising that individual subjects would differ.

2) If the results are not consistent within subject, then this is most likely due to the difference between PE and percent signal change, i.e. the denominator is different (the mean voxel intensity).

3) It may be that if this is an event-related design then the specifics of the regressor height estimation matter, in which case see Jeanette Mumford's web page on this.

Cheers.




On 15 Aug 2010, at 03:50, Elizabeth Losin wrote:

Hi FSLers,

I am comparing between several conditions within an ROI using featquery at the top level (mixed effects over 19 subjects) of a three level design (4 runs per subjct). The ROI was derived from an interaction contrast so I already know I have significant differences within this region at the whole brain level, now I just want to make a graph to illustrate the relationship between the 4 conditions simultaniously.  Each subject sees each of the following four conditions and I have compared each to the same fixation baseline.

A>rest,  B>rest,  C>rest,  D>rest

When I extract mean percent signal change for these conditions they fall in one order.  For example: A(.68) > D(.61) > B(.58) > C(.57) and when I compare this order to the means of any of the single subject data points also given to my by featquery (mean time series, pe1, thresh_zstat1) the order is different, for example for pe1: A(186.4) > C(164.4) > B(163.5) > D(86.5)  Here you can see that using the mean percent signal change values, condition D falls between A and B&C.   Whereas when I use pe1, condition D falls well below all three other conditions being about half of the next highest value.  How can this happen and which numbers can I trust?

Any help would be greatly appreciated.  Thanks.

Liz



---------------------------------------------------------------------------
Stephen M. Smith, Professor of Biomedical Engineering
Associate Director,  Oxford University FMRIB Centre

FMRIB, JR Hospital, Headington, Oxford  OX3 9DU, UK
+44 (0) 1865 222726  (fax 222717)
[log in to unmask]    http://www.fmrib.ox.ac.uk/~steve
---------------------------------------------------------------------------