Dear SPMers,
We previously posted a note about the fact that the regressors in SPM
'99 are centered on zero whereas they are not in SPM 2. It turned
out that this difference affected the orthogonality estimates between
conditions in the model, since the formula for calculating
orthogonality only reduces to a correlation coefficient when the
regressors have a mean of zero.
We have also noticed that the regressors are scaled differently in
SPM 2 compared with SPM '99. The regressors in SPM 2 are about 13
times as large as the corresponding values for the regressors in SPM
'99. This scaling difference markedly affects the efficiency values
we calculate for the contrasts in our event-related design.
Bottom line is that both efficiency estimates and orthogonality
estimates are different for SPM 2 relative to SPM '99, even though
the design we are using is identical.
Can someone explain why the regressors in SPM 2 are scaled to be so
much larger than they were in SPM '99, and also why they are no
longer centered on zero? (Will Penny suggested that the latter may
have simply been an oversight).
Thanks.
|