24
Conclusions from Previous Slides
¥ It is possible to find voxels where pre-whitening of different types (AR-only or MA-only or ARMA) is ÒoptimalÓ
H And voxels where pre-whitening makes little difference
¥ For many (most?) voxels, the pre-whitening details donÕt make a lot of difference in the statistics
H As long as something is done that is about right
H e.g., Using fixed AR(1) or MA(1) single parameter method was still OK-ish for single subject maps
oA few more extraneous small blobs
oBut fewer than pure OLSQ solution statistics
¥ Map of r1=correlation at neighboring TRs,
¥   as output by REML and ARMA(1,1) fit
H Same slice as previous slides (NIH 3 T data)
H In general, cortical gray matter shows more
H   correlation, but this result is not universal
H
H
0.0
0.7