Hello Shantanu:

You can plot the fit on top of the fMRI time series data using the NLfit
plugin. This will let you interactively change the signal model, parameter
constraints, etc. See Section 2 of 3dNLfim.ps for more details.

Another method is to (offline) generate the fitted time series for every voxel,
using the program 3dNLfim -snfit option. Then, in afni, use the Dataset #2
plugin to input the snfit 3d+time dataset, and from the Graph window select:
Opt -> Double Plot -> Overlay, and then select: Opt -> Tran 1D -> Dataset#2
This option is much faster, since it uses the pre-computed fitted time series.
However, it does not allow you to interactively change the model parameters.

The nonlinear regression method is inherently more difficult than the multiple
linear regression of 3dDeconvolve. To get a good fit for most voxels in
the dataset will require some experimentation with the parameter constraints,
and some insight into how the different model parameters effect the shape of
the fitted curve.

Doug Ward