Print

Print


Kelly,

To follow up on Steve's reply, if you've got paired (aka repeated measures)
data, then you *must *use a group file to tell randomise about what is
exchangeable.

If you have 3 subjects in group A and 2 subjects in group B, then your
design matrix might look like

   1     0     1     0     0     0     0     # Sick, Pre
  -1     0     1     0     0     0     0     # Sick, Post
   1     0     0     1     0     0     0     # Sick, Pre
  -1     0     0     1     0     0     0     # Sick, Post
   1     0     0     0     1     0     0     # Sick, Pre
  -1     0     0     0     1     0     0     # Sick, Post
   0     1     0     0     0     1     0     # Not Sick, Pre
   0    -1     0     0     0     1     0     # Not Sick, Post
   0     1     0     0     0     0     1     # Not Sick, Pre
   0    -1     0     0     0     0     1     # Not Sick, Post

and your contrast of interest [-1 1 ].  Randomise doesn't have to worry
about different group variance, but it *does* need to know that it only make
sense (under Ho) to swap pre and post data within subject.  Hence you need
to use a group file consisting of

1
1
2
2
3
3
4
4
5
5

which defines 5 'groups' (technically known as exchangeability blocks).  Can
you try this and let me know if works?

-Tom

____________________________________________
Thomas Nichols, PhD
Director, Modelling & Genetics
GlaxoSmithKline Clinical Imaging Centre

Senior Research Fellow
Oxford University FMRIB Centre



On Jan 24, 2008 10:10 PM, SUBSCRIBE FSL Kelly Brown <[log in to unmask]> wrote:

> Hi FSLers,
>
> I have a question, if I am using randomise to analyze an ANOVA type
> design,
> do I need to include a design.grp file for my two groups? I want to look
> at
> differences in pre and post and differences in sick and not sick... And if
> so, how should it look because I keep getting an error when I try to run
> randomise with the design.grp file and the design.fts file...I have
> checked
> the number of rows etc and they match my datapoints...
>
> group.datis not a valid vest file
> terminate called after throwing an instance of 'RBD_COMMON::BaseException'
> Aborted
>
> I get the same message when I try to make a F test file...
>
> design.grp
>
> %! VEST
> /NumWaves   1
> /NumPoints  10
> /Skip
> /PPheights  1     1
>
> /Matrix
> 1
> 1
> 1
> 2
> 2
> 1
> 1
> 1
> 2
> 2
>
>
> Thanks!
> Kelly Brown
> UCHSC Research Fellow
>
>