•Multicollinearity
Problem
→3dDeconvolve Error: Improper X
matrix (cannot invert X’X)
→X’X is singular (not invertible) ↔ at least one column of X is linearly dependent on the other columns
→normal
equation has no unique solution
→Simple
regression case:
•mistakenly provided at least two identical regressor
files, or some inclusive regressors,
in 3dDeconvolve
•all regressiors have to be orthogonal (exclusive) with
each other
•easy to fix: use 1dplot to
diagnose
→Deconvolution
case:
•mistakenly provided at least two identical stimulus
files, or some inclusive stimuli,
in 3dDeconvolve
Ø easy to fix: use 1dplot to diagnose
•intrinsic problem of experiment design: lack of
randomness in the stimuli
Øvarying number of lags may or may not help.
Ørunning RSFgen can help to avoid this
→see AFNI_data1/ht03/bad_stim/c20.bad_stim
•