Dear Wil,
you will probably want to model either main effects or 2-way-interactions,
not both. This is another example where SPM terminology is perhaps a bit
confusing, depending on your background knowledge. In short:
The design specification only determines which columns you include in your
design matrix. It does not do any tests for main effects or interactions
for you - this is done via contrasts.
The "interaction" columns are a generalisation of the "main effects"
columns, where each combination of levels from the factors involved is
modelled like a separate main effect. You can test for main effects of
one of the interaction factors using contrasts that average over the
levels of the other factor and include appropriate weights for the
modelled main effects as well.
For your design this would mean
1) Model "main effect subject" and "interaction between cond1 and cond2"
in your design matrix. This will set up N subject columns followed by
your 3x3 condition interactions.
2) To test main effects of one factor you would put a "1" over the columns
specifying the identical level of this factor - in your case this would be
three "1"s somewhere on the right in your contrast vector. Additionally,
you would weight each subject columns with 3/N (3 being the sum of
contrast weights over the conditions, N the #subjects).
3) For differential/parametric effects, the weights over the subject
columns usually cancel out.
Hope this helps,
Volkmar
On Tue, 24 Oct 2006, Wil Cunningham wrote:
> Hello,
>
> I am attempting to analyze a 3x3 within subjects ANOVA using SPM5.
>
> Using the flexible factorial setup, I created three factors
> 1. subject
> 2. cond1
> 3. cond2
> and then set up the subjects to have 9 images each (to complete the 3X3).
>
> Then, in the main effects & interactions section, I set up the three main
> effects, and the interaction between cond1 and cond2 {[2 3]}.
>
> Design1:
> When I estimate this, I can not generate any valid contrasts (e.g., 2 -1 -1;
> 0 1 -1 is not valid).
>
> Design2:
> If I drop the interaction, I can get the main effects.
>
> Design3:
> If I drop the main effects, I can estimate other contrasts (e.g., [2 2 2 -1
> -1 -1 -1 -1 -1; 0 0 0 1 1 1 -1 -1 -1]) for the first main effect.
>
> I am a little confused about how I should be running this. I would think
> that Design1 above is the way that I should do this, but is it perhaps
> Design3 -- and then I set up the main effects in the contrast editor?
>
> Any help would be greatly appreciated.
>
> Thanks in advance,
> Wil
>
>
--
Volkmar Glauche
-
Department of Neurology [log in to unmask]
Universitaetsklinikum Freiburg Phone 49(0)761-270-5331
Breisacher Str. 64 Fax 49(0)761-270-5416
79106 Freiburg http://fbi.uniklinik-freiburg.de/
|