¥ The tool for expressing an unknown function as a finite set of numbers that
can be fit via linear regression is an expansion in basis functions
H The basis functions yq(t ) are known, as is the expansion order p
H The unknowns to be found (in each voxel)
comprises the set of
weights bq for each yq(t )
¥ Since b weights
appear only by multiplying known values, and HRF only
appears by in final signal model by linear
convolution, resulting signal model is still solvable by
linear regression