Hello Rutger:

The coefficient output from 3dDeconvolve represents the amplitude of the
response relative to baseline, NOT the percent change relative to baseline.

Therefore, if you want to use percent change, you will have to perform some
additional calculation (using 3dcalc, for example).

Program 3dDeconvolve estimates a single measurement variance. The calculation
assumes that the variance is constant across the entire time series. This
has implications for the analysis, depending on how changes in the baseline
effect the fMRI data.

If a change in the baseline between runs has no effect on the signal or noise
amplitudes, then no further adjustment is required.

If you believe that a change in baseline between runs results in a proportional
scaling of the amplitude of the response, as well as the measurement error,
then it would be better to "undo" this scale factor prior to the 3dDeconvolve
analysis. As I mentioned in the previous message, you could convert the
individual run time series to % signal relative to the baseline for each
individual run, prior to concatenation (and prior to 3dDeconvolve analysis).
In this case, the 3dDeconvolve output coefficients will represent % change
relative to baseline.

Doug Ward