Hi Javier,

I'm very sorry -- your email got lost in my pile, and I forgot about 
it! Did anyone else get back to you about this? And/or did you manage 
to work it out yourself?

If not, in rough terms, I think you want to specify flexible factorial 
and include a factor for subject, as well as your other factors. Then 
when you specify scans (or "specify all" and a factor matrix) you can 
denote the levels of the other factors for each subject.

So e.g. for "specify all", the factor matrix might have a first column 
of all ones for replication, then a column with blocks of four numbers 
(1 1 1 1 2 2 2 2 ... 9 9 9 9) for each subject, and then columns for 
the levels of the other two factors (e.g. 1 1 2 2 1 1 2 2 ... and 1 2 
1 2 1 2 1 2 ...).

It's possibly simpler to treat the 2*2 factorial as a 1*4 if you don't 
need different non-sphericity options for the different factors. [Or 
maybe it is genuinely 1*4? -- I'm not sure I completely understood 
your original description...] In that case you would have just the 
subject factor and the other factor, with its column in the factor 
matrix being (1 2 3 4 1 2 3 4 ...). This should match the design 
matrix that I suggested before:
   [kron(ones(4,1), eye(9)) kron(eye(4), ones(9,1))]

Does that help? Or, if not, there have been a few more recent posts on 
this, perhaps some of those have helped / will help?

Sorry for the very slow reply,

Javier Gonzalez-Castillo wrote:
> Hello Ged,
>  Thanks very much for your answer. And now sorry for what I imagine is a
> very simple question: how do I define the within-subject ANOVA model in SPM5
> so I get a design matrix such as the one you propose (X =
> [kron(ones(4,1), eye(9)) kron(eye(4), ones(9,1))] )
> What second-level model type should I use: Full Factorial, Flexible
> Factorial? Could you please give me some fast guidelines about how to define
> the model the way you propose. I've tried different approaches, but I
> definitelly don't know how to do it.Do I have to do something different at
> the first level or I just use the same con files I was using in my previous
> approach?
> Thanks again,
> Javier Gonzalez-Castillo
> PhD Student - Biomedical Engineering - Purdue University