Posted by Michael S. Beauchamp on November 15, 2002 at 15:11:29:
In Reply to: how to calculate the area under the curve posted by Serge Rombouts on November 15, 2002 at 11:01:08:
Dear Serge,
You are on the right track.
The main reason to normalize by the baseline is because this converts your measure to "percent signal MR change", which is a good unit for comparing across subjects and studies (for more information on different ways to perform the normalization, there have been a number of message board posts from me and others on the topic). If you don't convert to percent signal change, changes in the baseline will artificially add variance to your ANOVA.
e.g. subject 1, signal in hippocampus goes from 1000 to 1050 = 50 irf units = 5% signal change
subject 2, signal in hippocampus goes from 500 to 525 = 25 irf units = 5% signal change
If I run ANOVA on 25 and 50, there will be more subject variance than if I run on percent signal change (5 and 5).
re: area under the curve
It is difficult to conceive of an impulse response function with only 2 points. To properly create an irf, you would like to have more points, say 15 with 1 second resolution. In this case, simply adding the values at each point gives a good approximation of the area under the curve;
e.g. [
ugrad.math.ubc.ca]
For slightly better results, we might only consider the area under the main body of the irf, for instance from 2-8 seconds.