Dear Mike,
>> I'm looking at residuals and their effect on varcopes, and would like to look at the keff information (described in tech note TR01MW1, eq 3 among other places). In looking at the varcope as an end result, I see that while the sigmasquareds are provided as a file in the stats directory that the keff values do not seem to be.
>> I had hoped that the neff file contained this information, but the neff values don't seem to work out.
>> I understand that it is possible to reconstruct keff from the autocorrelation values in the threshac file, from filtering info and design.mat, but FSL has already performed that calculation and I hoped to avoid it.
>> Is there a file, or combination of files, that can provide this information while taking advantage of calculation already performed?
Yes. It is in the neff file.
neff is 1/keff, where keff is given in eqn 3 in:
Woolrich et al. Temporal autocorrelation in univariate linear modeling of FMRI data. Neuroimage (2001) vol. 14 (6) pp. 1370-86
i.e. var(COPE)=keff *sigma^2
>> On a related note, the images in the threshac file increased in the last few years, it seems (at least, in an early project I have 1 image while in a later project I have 28, for the same data). Is there comment on this in a release note, or can you comment briefly on the change?
That is unexpected behaviour for the threshac to only have 1 value. The only reason I can think this would have happened is if you had turned off prewhitening.
Cheers, Mark.
On 2 Jul 2012, at 17:26, Michael Rohan wrote:
> Hi,
> I'm looking at residuals and their effect on varcopes, and would like to look at the keff information (described in tech note TR01MW1, eq 3 among other places). In looking at the varcope as an end result, I see that while the sigmasquareds are provided as a file in the stats directory that the keff values do not seem to be. I had hoped that the neff file contained this information, but the neff values don't seem to work out. I understand that it is possible to reconstruct keff from the autocorrelation values in the threshac file, from filtering info and design.mat, but FSL has already performed that calculation and I hoped to avoid it.
>
> Is there a file, or combination of files, that can provide this information while taking advantage of calculation already performed?
>
> On a related note, the images in the threshac file increased in the last few years, it seems (at least, in an early project I have 1 image while in a later project I have 28, for the same data). Is there comment on this in a release note, or can you comment briefly on the change?
>
> Thanks for any info,
> Mike Rohan
> McLean Hospital
|