PowerPoint Presentation - Time Series Analysis in AFNI

–55–

Deconvolution Signal Models

• Simple or Fixed-shape regression:

H We fixed the shape of the HRF

H Used waver to generate the signal model from the stimulus timing

H Found the amplitude of the signal model in each voxel

• Deconvolution or Variable-shape regression:

H We allow the shape of the HRF to vary in each voxel, for each stimulus class

H Appropriate when you don’t want to over-constrain the solution by assuming an HRF shape

H Caveat: need to have enough time points during the HRF in order to resolve its shape