GBefore averaging the
values in each voxel, one option is to remove any ÒuninterestingÓ time lags and average only those time
lag points that matter
åIn this example, ÒuninterestingÓ refers to time lags
whose percent signal change value is close to
zero
åIf we included these uninteresting time lags in our mean
calculation, they would pull the mean
closer to zero, making it more difficult to pick up any statistically significant results in the ANOVA
GIn this example, the
IRF is fairly small from time lags 0-3, but begins to spike up at time lag 4.
It peaks at time lag 7, declines at 9, drops sharply by 10, and decreases steadily from there.
åGiven this information, one option
is to compute the mean percent
signal change for time lags 4-9
only