Hi,
In the context of regressing out motion effects, it is really simple - just a linear offset and scaling transformation. As long as the GLM takes care of the mean component (either by demeaning quantities, or including a constant regressor) then the scaling part is compensated for by the beta parameter in the GLM (which is scaling the regressor).
I hope this helps.
All the best,
Mark
On 27 Mar 2014, at 15:07, "Watson, Christopher" <[log in to unmask]> wrote:
> In general (not specific to neuroimaging) it is not necessary to transform both the response and predictor variables.
> e.g. Using log(Y) is just as acceptable as using Y; although your interpretation will have to change.
>
> ________________________________________
> From: FSL - FMRIB's Software Library [[log in to unmask]] on behalf of Andreas Werner [[log in to unmask]]
> Sent: Thursday, March 27, 2014 10:51 AM
> To: [log in to unmask]
> Subject: [FSL] Aw: Re: [FSL] fix - regress out motion confound
>
> Thank you for your quick reply.
> --It's not important to......to the extent that I don't really understand where the question is coming from ;-)
> I am studying this script to learn how regressing out confounds is professionally done and implemented as this is a processing step often needed.
> In the mentioned script, z-transformed motion confounds are agressively regressed out of a) the z-transformed melodic_mix matrix and b) the untransformed filtered_func_data file.
> I was curious to know, why z-transformation was not applied to filtered_func_data, as I presumed that in a GLM(Y=X*b) both X(=confounds) and Y(=melodic_mix or filtered_func_data) must be transformed to stay consistent.
> As I understand this is is not important in the context of regressing out.
> Thank you for pointing that out!
> Cheers.
>
> Gesendet: Donnerstag, 27. März 2014 um 14:39 Uhr
> Von: "Stephen Smith" <[log in to unmask]>
> An: [log in to unmask]
> Betreff: Re: [FSL] fix - regress out motion confound
> Hi
>
> On 27 Mar 2014, at 13:22, Andreas Werner <[log in to unmask]> wrote:
>
> Dear Experts,
> studying the FIX scripts (fix_3_clean.m) I noted that motion parameter confounds are z-transformed (mean removed and normalized to unit variance aross time) before they are regressed out.
>
> maybe - but that will have no effect on their regression out of the data timeseries - that linear transformation doesn't affect the "space they span", in GLM lingo.
>
> Melodoc_mix is also z-transformed, but the
> filtered_func_data is not. Why ?
>
> It's not important to......to the extent that I don't really understand where the question is coming from ;-)
>
> Cheers.
>
>
> Do you generally recommend to z-transform motion parameters before regression rather than just use the untransformed entries in prefiltered_func_data_mcf.par ?
> Cheers,
> Andreas
>
>
> ---------------------------------------------------------------------------
> Stephen M. Smith, Professor of Biomedical Engineering
> Associate Director, Oxford University FMRIB Centre
>
> FMRIB, JR Hospital, Headington, Oxford OX3 9DU, UK
> +44 (0) 1865 222726 (fax 222717)
> [log in to unmask] http://www.fmrib.ox.ac.uk/~steve
> ---------------------------------------------------------------------------
>
> Stop the cultural destruction of Tibet<http://smithinks.net>
|