Hi,
No, this model is indeed most strictly accurate for 8 subjects (or 1) as
there is no repeated measures aspect in this particular model.
The reason why you don't want (and in fact can't use, in an f-test) a 4th
contrast (D - mean) is that the space spanned by the cells is already
fully spanned by the first three contrasts because of the demeaning. So
you are asking the question of whether there is any effect from any of
them. Imagine if you had just three levels, represented by 3 points in
x,y,z space; if you demean their values so that they are based around the
origin (and that the origin is now contained in the plane through the 3),
then vectors from the origin pointing towards 2 of them define the plane
containing all three and so fully "span that space" (i.e. that plane). So
you don't need to use the 3rd - because of the demeaning it doesn't add
anything. What the f-test is doing is projecting the data onto that plane
to see how much variance is explained in it.
Cheers, Steve.
On Fri, 12 Nov 2004, Ramapriyan Pratiwadi wrote:
> hi all -
>
> i am seeking a way to see how 4 levels of 1 factor among multiple
> subjects are different from one another,(where A is different from
> {B,C,D}, B from {A,C,D}, etc...)
>
> I believe this is the ANOVA: 1-factor, 4 levels(repeated measures) as
> described in the FEAT webpage.
>
> ( as a side note, isn't the ANOVA: 1 factor, 4 levels graphic on the
> webpage really for 2 subjects and not 8 as stated in the paragraph? )
>
> i *think* i understand how the analysis works. however, i don't see why
> a fourth contrast is not necessary; for levels A,B,C,D, don't we want to
> know if each is different from the mean(A,B,C,D) ? with the method
> described on the webpage, as i understand it, i only get A vs.
> mean(A,B,C,D), B vs. mean(A,B,C,D), and C vs. mean(A,B,C,D).
>
> thoughts?
>
> thanks in advance.
>
> -ram
>
> ---------------------------------------------------------------------------
> Ram Pratiwadi
> Research Assistant, University of Pennsylvania
> [log in to unmask]
>
> --
> This message has been scanned for viruses and
> dangerous content by MailScanner, and is
> believed to be clean.
> MailScanner thanks transtec Computers for their support.
>
Stephen M. Smith DPhil
Associate Director, FMRIB and Analysis Research Coordinator
Oxford University Centre for Functional MRI of the Brain
John Radcliffe Hospital, Headington, Oxford OX3 9DU, UK
+44 (0) 1865 222726 (fax 222717)
[log in to unmask] http://www.fmrib.ox.ac.uk/~steve
|