Dear Federico,
actually, Rik Henson pointed out that my reply was not quite correct in
statistical terms. SPM uses the factor-wise error covariance specification
to model a single, non-spherical error process at each level of
statistics. If you want to take into account only some of your sources of
error, you will have to use different levels of analysis by taking con-
(i.e. effect size) images from a number of analyses into a new GLM.
Hope, I got this right now...
Volkmar
On Mon, 5 Mar 2007, Federico Tubaldi wrote:
> Thank you for your support.
>
> Federico.
>
> -----Messaggio originale-----
> Da: SPM (Statistical Parametric Mapping) [mailto:[log in to unmask]] Per
> conto di Volkmar Glauche
> Inviato: luned́ 5 marzo 2007 10.13
> A: [log in to unmask]
> Oggetto: Re: [SPM] POOLED and PARTITIONED error in SPM5
>
> Dear Federico,
>
> you are absolutely right in your conclusions: The design matrix for both
> models will be identical, but the 2-factor model is more general in terms of
> error partitioning. You can still pool your error, if you specify equal
> variance and indepence for all factors and levels, but you can also
> partition your error for either factor a or b or even both factors.
>
> Volkmar
>
> On Sat, 3 Mar 2007, Federico Tubaldi wrote:
>
>> Dear Experts,
>>
>> In full factorial design option is it different to specify one factor
>> with four levels (a1b1; a1b2; a2b1; a2b2) or two factors (A,B) with
>> two levels
>> (a1,a2 and b1,b2)? I think that in the former situation I have a
>> pooled error as when I used one-way anova within subjects in spm2 (spm
>> does not know the factorial structure) while in the latter situation I
>> have a partitioned error (spm knows the factorial structure). Is this
> correct?
>> I have ran the two models: the design matrixes are the same and the
>> results are strongly similar. Is this plausible or should I get some
>> differences for example in matrix structure?
>>
>> Thank you.
>>
>> Federico.
>>
>>
>>
>
> --
> 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/
>
> --
> No virus found in this incoming message.
> Checked by AVG Free Edition.
> Version: 7.5.446 / Virus Database: 268.18.6/708 - Release Date: 02/03/2007
> 16.19
>
>
>
--
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/
|