The -noest option in film_gls turns OFF the prewhitening (in which case
the "-sa -ms 5" arguments are irrelevant and superfluous). The Thousand
Functional Connectomes command that you mentioned could have generated
basically identical residuals (up to a small difference in the size of
the mask) using:
fsl_glm -i input_data -m mask -d design.mat --demean --out_res=res4d
cheers,
-MH
On Fri, 2012-03-23 at 12:43 -0500, Christopher Bell wrote:
> Does anyone have any comment on why prewhitening would not be
> recommended for seed analysis?
> We are using the Thousand Functional Connectome for our pre-processing
> and it uses feat_model for orthogonilization and regression
> of WM,CSF,WB and motion time-series. It also then calls film_gls with
> the parameters below.
>
> Are there pros/cons to this method vs. simply using fsl_glm, which is
> what I have done in the past. I think the results
> end up being pretty different, so it seems important. If someone could
> elaborate on exactly what this film_gls call is doing I would
> appreciate it,
> is it simply prewhitening?
>
> ## 7. Get residuals
> echo "Running film to get residuals"
> film_gls -rn ${nuisance_dir}/stats -noest -sa -ms 5
> ${func_dir}/${rest}_pp.nii.gz ${nuisance_dir}/nuisance.mat ${minVal}
>
>
> Chris
>
> On Thu, Mar 22, 2012 at 2:13 PM, Christian F. Beckmann
> <[log in to unmask]> wrote:
> Hi
>
> running all this through FEAT is really not the best option -
> while it could be tweaked to work, I think you should instead
> run this via fsl_glm and randomise. You can then continue to
> use res4D
> hth
> Christian
>
> On 22 Mar 2012, at 10:35, Dr. David Watson wrote:
>
> > Currently I am clear as far as extracting the Res4D image
> normalising this (with the appropriate brain mask) and
> extracting the seed time series for the VOI. I have checked
> these stages and everything looks okay. It is when I put the
> normalised Res4D image and the selected ts as an EV back into
> FEAT that I'm confused. What should be on and off in the
> settings. For example in Pre-stats I do not select Motion
> Correction or Slice Timing but do select BET brain extraction;
> In Stats I have Pre-whitening selected, put in the ts as an EV
> but have Motion Parameters deselected; I leave Post Stats at
> their defaults. Does this sound a reasonable setup?
> >
> > David
>
>
|