> If I convert my time series to % signal change (a(t)/mean(a)*100)
> will the values of the regression coefficents returned by
> 3dDeconvolve reflect % signal change values?
Yes, usually you can interpret those regression coefficients as percent signal change if the regressors are scaled properly.
> If I did not scale the data prior to 3dDeconvolve how could I obtain
> % signal change values from the regression coefficents?
If there is only one run of data, you can estimate % signal change by dividing those regression coefficients by the baseline coefficients at each voxel. However with multiple runs of data unscaled, regression analysis with combined data could be problematic because of baseline variation across runs.
> I have several non-linear drifts due to a long run and irregularities
> in scanner function. Should I attempt to correct these drifts using
> a high pass filter implemented using 3dFourier or can I detrend
> the time series within 3dDeconvolve itself?
Either is probably fine although detrending in 3dDeconvolve is preferred. Or you can include those frequencies as regressors in 3dDeconvovle.
Gang