Posted by Serge Rombouts on November 15, 2002 at 11:01:08:

Hi,

I want to calculate the area under the impulse response function (IRF) curve following Example 1.4.2.5 on page 32 of 3dDeconvolve.ps and use these measures for a 3dANOVA2 analysis.
I have used (first level):
3dDeconvolve \
....
-polort 1 \
-stim_file 1 activation.1D \
-stim_minlag 1 0 -stim_maxlag 1 1 \
-iresp 1 bill.irf \
-fout -rout -tout -bucket bill.bucket

In my opinion it is not really possible to get the area under the curve, but rather an average of the least squares estimates of the response amplitudes at different time lags (in my case: (bill.irf[0]+bill.irf[1])/(2) ).
However, it is stated in the example that one should divide by bill.bucket[0] to normalize this measure to baseline (that is, divide by the constant term). Is this an error in the documentation or am I misinterpreting this?
Is my approach correct to simply calculate (bill.irf[0]+bill.irf[1])/(2) and then use this in 3dANOVA2 to compare between subjects?

Thank you,
Serge.