Hi all,
I was hoping someone could shed some light on the inconsistent contrast
values I've been getting for my data. It's an old data set from a 1.5T
scanner, preprocessed in SPM99 and analyzed in SPM2. The problem does not
seem to be the switching between the 2 SPMs since I've run through all the
steps in SPM2 for one subject, and I still get similar contrast values.
Using MarsBaR to combine activation ROIs with anatomical ROIs, I output a
statistic table which produced contrast values ranging from 411.46 to 6.44
(but my t-statistics are all in the range of 3.14 to 1.65). Below is an
example from one subject for two ROIs illustrating the wide range of values:
ROI name: Contrast value: t statistic: Uncorrected P: Corrected P
------------------------------------------
memimagvsrestLparahipp: 411.10: 1.92: 0.028176: 0.028176
imaginevsrestRhipp: 6.44: 2.14: 0.016536: 0.016536
This does not, however, seem to be a MarsBaR-specific problem since plotting
the contrast estimates in SPM yields similarly high values.
I was hoping that someone could tell me whether or not these contrast values
are at all reasonable, and if they aren't, what could possibly account for
such high values. Previous work in our lab as well as all the literature
I've come across seems to put the values in a range between 10 and -2.
Could someone also explain in more detail exactly what the contrast value
calculation entails? Is it merely the numerator of a T-statistic?
Thank you for your help,
Christine Cox
Cognition and Neuroimaging Labs
University of Arizona
|