Hi,
On 25 Oct 2008, at 23:21, Geoff Kerchner wrote:
> Dear all,
>
> I am a newcomer to linear regression matrix design, so apologies for
> asking
> some basic questions. I want to use randomise to analyse DTI data
> processed
> through tbss, and I'm using the Glm GUI to create my design.met and
> design.con files.
>
> My study uses a single group of subjects, each with several sets of
> behavioral
> data (reaction times to various tasks). I am hypothesizing that FA
> values will
> correlate negatively with RT. Here are my questions:
>
> (1) Should I run each set of RT measures separately, or can this be
> achieved
> in a single matrix design? I am not interested in interactions
> between these
> different measures, only the possible negative correlation between
> each
> measure and FA values.
You can do either - depending on whether you want to be more
conservative (putting them all together in one design) or more liberal
(one design and run of randomise for each covariate separately, where
it can 'take all of the variance related to it').
> (2) My understanding is that the design matrix will have a column of
> 1s to
> indicate one group, and then subsequent columns containing RT data. My
> understanding also, from reading other posts, is that RT data should
> be
> demeaned.
>
> [1 rt_subj1]
> [1 rt_subj2]
> [1 rt_subj3]
> [1 rt_subj4]
> ...
That's right.
> (3) To set up the contrast mask in the case of a simple regression
> between
> FA and on set of RT data, would it just be [-1] (for a negative
> correlation)?
Yes, except that you also need a contrast entry for the constant
height regressor, i.e. [0 1] or [0 -1].
> (4) How can I adjust for age or other factors? I would include these
> as
> additional explanatory variables, but I'm not sure how to set up the
> contrast
> matrix.
You can just add these as further (demeaned) regressors if you wish,
giving confound regressors '0' in the contrast.
Cheers.
>
>
> Thanks in advance for your help,
>
> Geoff
>
---------------------------------------------------------------------------
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
---------------------------------------------------------------------------
|