dear SPM group,
here a naive question about a group analysis of fMRI data over which
we - some people in our fMRI groups - strain our brains for a few days
now. the main question is how to generate contrasts on the first
level (as INDEPENDENT measures vs as ONE mean-contrast per subject),
and how to process those constrasts on the second level for the
following admittedly simple scenario:
2 groups, g1 & g2, each with 13 subjects.
2 sessions A & B per subject (sometimes just one session).
A & B are not identical as order of stimuli is different
stimuli are regular tones & mismatch tones (rare)
focus of interest on effect of mismatch tones
analysis 1 (ONE contrast):
fixed effect analysis:
together for both sessions A & B, treating sessions A & B as
ONE contrast:
- first regressor for effect of tones
- second regressor for effect of mismatch tones (parametric)
- creating one subject specific constrast file for a) effect
of tones and for b) effect of mismatch for both sessions A &
B together.
in contrast manager: [1 0 1 0 0] and [0 1 0 1 0].
random effect analysis RFX:
- for simple groups effects: pooling 13 subject specific
contrast files of group g1 or g2 for 1-sample t-test
- for group comparison: pooling 26 (g1 & g2) subject specific
contrast files for 2-sample t-test
analysis 2 (INDEPENDENT contrasts):
fixed effect analysis:
separately for sessions A & B, treating sessions as
INDEPENDENT measures:
- first regressor for effect of tones
- second regressor for effect of mismatch tones (parametric)
- creation of two session specific contrast files for a)
effect of tones and for b) effect of mismatch, for A or B
independently.
in contrast manager: A: [1 0 0] and [0 1 0]
B: [1 0 0] and [0 1 0]
random effect analysis RFX:
- for simple groups effects: pooling 26 contrast files for
sessions A & B for group g1 or g2.
- for group comparison: pooling 52 contrast files for sessions
A & B for both groups g1 & g2.
which analysis method should we prefer?
thanks in advance for any input.
pisti
-----
[log in to unmask]
|