¥ Between OLSQ and GLSQ+REML:
H Individual
subject thresholded activation maps may
differ very little, some, or a lot
¥ Level of temporal correlation determines
how much difference GLSQ makes to individual subject statistics
H Amount of temporal correlation seems to
depend on magnetic field strength, other scanner
details, pulse sequence, É
H Effect of temporal correlation also
seems to depend on stimulus timing
H As theory indicates:
o Temporal
correlation means noise variance depends on frequency
o So amount of noise that interferes with (Òlooks likeÓ) the
signal will depend on frequencies at
which the hemodynamic response is appreciable
¥ Next slides: Group activation maps,
GLSQ+REML vs OLSQ
H 2 cases from NIH: Event-related and
Block:30s designs
H DonÕt have enough FBIRN-1 subjects to do
a group analysis