> I am doing a simple block design experiment where I am have
> two conditions faces and houses. Both face and house stimuli
> will parametrically vary in luminance.
>
> I would like to use the first order polynomial basis function
> to see whether there are any areas which are modulated by
> changes in luminance of faces, but not houses.
>
> My main question is how to define the contrast?
>
> >From what I understand my GLM looks like this:
>
> Y = Faces*beta1 + (Faces*Time)*beta2 + Houses*beta3
> (Houses*Time)*beta4 + mean*beta5 + error
>
>
> Would it be correct if I defined my contrast as [0 1 0 -1 ]
> and used a t-test?
Yes, although this contrast will not show voxels that are
modulated by face luminance but NOT house luminance. The
latter kind of claim is difficult with classical statistics.
Rather, this contrast will show voxels where the (linear)
effect of luminance is GREATER for faces than for houses.
This is probably what you want (but if you're sure you want
to test your hypothesis as originally stated, then you could
test [0 1 0 0] and exclusively mask with [0 0 0 1], but you
will be trying to "prove" the null hypothesis - you could
try a PPM instead!)
> How can I decide if I need to use a higher order basis
> function and which order to use?
If you have N levels of luminance, you can model up to (N-1)
polynomial orders. If you have NO idea what form the function
relating BOLD to luminance takes (e.g, even whether or not it
is monotonic), and assuming you have >1 replications per level,
there is little harm in modelling all N-1 orders, since they
are orthogonal to each other. (If you have pilot data, you could
test each order separately, from N-1 to 1, until you find a
reliable effect, and then remove from future models orders
higher than that.) If you expect a certain form however (eg linear),
then stick with just the relevant order (eg 1 in this case), else
your model can be too flexible (see:
http://www.jiscmail.ac.uk/cgi-bin/wa.exe?A2=ind0504&L=spm&P=R3153&I=-1
)
Rik
--------------------------------------------------------
DR RICHARD HENSON
MRC Cognition & Brain Sciences Unit
15 Chaucer Road, Cambridge,
CB2 2EF England
EMAIL: [log in to unmask]
URL: http://www.mrc-cbu.cam.ac.uk/~rik.henson
TEL +44 (0)1223 355 294 x522
FAX +44 (0)1223 359 062
MOB +44 (0)794 1377 345
--------------------------------------------------------
|