Hi,
So - you want to exclude the _mean_ effect of age in the two groups,
and compare how FA varies with age within the two groups separately?
This all looks fine, but if you want to compare the two groups'
variation of FA with age, you need to test that directly with a [1
-1] or [-1 1] contrast - it's never safe to show that different
thresholded maps look different, as one could be just above threshold
and the other just below, and they might not be significantly different.
Cheers.
On 11 Jun 2007, at 19:51, Ping-Hong Yeh wrote:
> I am trying to understand the 'regress out' using randomize and
> need your
> help on clarification of my approach. Supposedly the hypothesis is
> that age
> has different effect on FA between groups.
> To test the hypothesis, age is in the design matrix and demeaned
> between groups
>
> EV1 EV2
> -12.5144662 0
> -2.2267942 0
> 13.9978628 0
> 0 -9.5596513
> 0 -10.0473223
> 0 -0.2555413
> -13.6678902 0
> 5.0526578 0
> -3.9555622 0
> -1.1561092 0
> -2.1692602 0
> 2.9238908 0
> 0 0.0485677
> 0 9.4184307
> 0 -7.1322543
> 0 9.7198007
> 0 5.0787047
> 0 5.6814447
> 13.7156708 0
> 0 -8.2993773
> 0 5.3471977
>
>
> The group is in the confound matrix
> 1 0
> 1 0
> 1 0
> 0 1
> 0 1
> 0 1
> 1 0
> 1 0
> 1 0
> 1 0
> 1 0
> 1 0
> 0 1
> 0 1
> 0 1
> 0 1
> 0 1
> 0 1
> 1 0
> 0 1
> 0 1
>
> , and the contrast matrix is
> 1 0
> -1 0
> 0 1
> 0 -1
>
> The results of suprathreshold cluster tests showed age is
> significantly
> inversed with FA in clusters over different region in patient
> group, but not
> in control. Can I draw the conclusion that age has more (or
> different)
> inverse effect in FA in patient than controls? Is following
> mathematical
> expression is what randomize does according to above design matrix?
> That is considering
>
> FA = lambda 1 + lamda2*group
>
> To control the possible effect of age effect between groups,
> FA = lambda 1 + lamda2*group +beta*age
>
> Then the test if age is significant, that is,
> Ho: beta=0 vs H1: beta not equals zero.
>
> What not clear to me are those orthogonal EVs, demeaned age between
> groups,
> which are not usually used in generic GLM. How would this approach
> different from fitting linear regression models , (without
> demeaning) to
> each group separately for evaluating the age effect ?
>
> Best,
> Ping-Hong Yeh
------------------------------------------------------------------------
---
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
------------------------------------------------------------------------
---
|