Francesca Sabattoli TISCALI wrote:
> Dear SPM users,
>
> could someone explain me the meaning of "mean-centering the covariates
> or nuisance (no centering or around overall mean)"? Thanks
>
>
>
> Francesca
If x is a user-specified covariate - like Reaction Time (RT) -
ie. a column in the design matrix X, then
mean centering around the overall mean implies the MATLAB operation
> x=x-mean(x)
ie. mean(x) computes the overall mean (ie. over all entries in X)
this is then subtracted from all elements.
But, if you're looking for factor by covariate *interactions*
(if you're not you can stop reading now !) you may wish to do a different mean operation.
Eg. if your factor is experimental condition, you may have a set of RTs for each condition.
It may be the case that BOLD is positively correlated with RT in condition 1 but
negatively correlated in condition 2 ie. there is a condition-by-RT
interaction. To find this out you don't want your interaction
regressors to contain the main effects of condition. Therefore, you'd
center your data, in this exampe, around condition means.
See text entries below for further info.
Hope this helps.
Best,
Will.
From spm_spm_ui.m:
% * Covariate centering: At this stage may also be offered "covariate
% centering" options. The default is usually that appropriate for the
% interaction chosen, and ensures that main effects of the interacting
% factor aren't affected by the covariate. You are advised to choose
% the default, unless you have other modelling considerations. The full
% list of possible options is:
% - around overall mean
% - around replication means
% - around condition means (across group)
% - around subject means (across group)
% - around group means
% - around condition means (within group)
% - around subject means (within group)
% - <no centering>
>
>
>
> Laboratory of Epidemiology & Neuroimaging
> IRCCS San Giovanni di Dio FBF - The National Center for Research and
> Care of Alzheimer's Disease
> via Pilastroni 4, 25125 - Brescia, Italy
> Tel: +39 030 3501 361, Fax: +39 02 700 435 519 (preferred) or +39 030
> 3533 513, www.centroAlzheimer.it <http://www.centroAlzheimer.it>
>
--
William D. Penny
Wellcome Department of Imaging Neuroscience
University College London
12 Queen Square
London WC1N 3BG
Tel: 020 7833 7475
FAX: 020 7813 1420
Email: [log in to unmask]
URL: http://www.fil.ion.ucl.ac.uk/~wpenny/
|