6
•Regression Model (General Linear System)
→Simple Regression Model (one regressor): Y(t) = a0+a1t+b r(t)+e(t)
•Run 3dDeconvolve with regressor r(t), a time series IRF
→Deconvolution and Regression Model (one stimulus with a lag of p TR’s):  
• Y(t) = a0+a1t+b0f(t)+b1f(t-TR)…+bpf(t-p*TR)+e(t)
•Run 3dDeconvolve with stimulus files (containing 0’s and 1’s)
•Model in Matrix Format: Y = Xb + e
→X: design matrix - more rows (TR’s) than columns (baseline parameters + beta weights).          
→           a0     a1       b                                       a0    a1     b0    …    bp
•             ------------------                      ------------------------
q1     1     r(0)                         1   p     fp     …    f0 
q1     2     r(1)                                      1   p+1 fp+1 … f1
q    .  .  .         .  .  .
q1   N-1    r(N-1)                                1   N-1 fN-1 … fN-p-1
→
→e: random (system) error N(0, s2)
•
•
Basics about Regression