Hi All,
We've recently run into some numerical overflow problems when reading our AFNI data into Matlab.
Initially all of the EP subbriks are shorts and the scale seems to be around 0 to 800. When read into Matlab the values range from -32767 to 32767, but there seems to be some wrapping when we display the data. (Values that are over 32767 are wrapped back around as negative values?)
When we write these datasets back out into BRIK and HEAD format AFNI seems to read them without any of the wrapping that we see in Matlab.
How is this wrapping corrected by AFNI?
Also, a 3dinfo on a regular EP dataset produces the following:
-- At sub-brick #680 '#680' datum type is short: 0 to 742
A 3dinfo on the output of 3dDeconvolve:
-- At sub-brick #34 'L1-2_acheive[0] Coef' datum type is short: -32767 to 16996 [internal]
[* 0.000403292] -13.2147 to 6.85435 [scaled]
The "internal" values are what we seem to be seeing when reading the regular EP dataset into matlab. Is there any way to see the internal values on a pre-deconvolution EP dataset?
Emmette