Just a short note to say that this problem has been resolved thanks to the
help from Volkmar (and others back-channel).
The problem arose because the estimation of covariances for the two groups
of subjects was vastly different. Group1 had a huge error variance while
it was small in Group2. This was a problem not with SPM5 or version of Matlab
or the groups themselves, but with how I had originally estimated GLMs / con*imgs for Group1.
Somehow an error had crept in only for Group1 but not for Group2.
When this was resolved (redo the GLMs for Group1), the other problem with
the fullfactorial design resolved too.
Fatima
On Mon, 14 May 2007, Volkmar Glauche wrote:
> Dear Fatima,
>
> can you please send me the batch definition file from SPM5 job manager? There
> must be something strange going on...
>
> Volkmar
>
> On Fri, 11 May 2007, FATIMA HUSAIN wrote:
>
>> No, I cannot set up the contrasts manually.
>> Every contrast entered is deemed "invalid". The design matrix is not set up
>> correctly in that only 1 group's data seem to be present (the smaller
>> group). If the design matrix is reviewd before the estimation step, it
>> appears to be correct.
>>
>> Thanks,
>>
>> Fatima
>>
>> On Fri, 11 May 2007, Volkmar Glauche wrote:
>>
>>> Dear Fatima,
>>>
>>> it looks as if the error occured _after_ estimation (e.g. in a contrast
>>> manager batch). Are you able to view results for the analysis if you set
>>> up contrasts manually?
>>>
>>> Volkmar
>>>
>>> On Thu, 10 May 2007, FATIMA HUSAIN wrote:
>>>
>>> > I am trying to run a simple fullfactorial (mixed ANOVA) 2nd-level
>>> design > in SPM5. (this is because I could not do this in SPM2 in a >
>>> straighforward manner).
>>> > Factor1= group, levels =2, independent, unequal variance
>>> > Factor2= task, levels=2, dependent, equal variance.
>>> > > Group1 has 8 patients, Group2 has 16 controls.
>>> > > Should be easy right? But nooo...
>>> > > I get an error during estimation (appended below). If I change
>>> Factor1's > variance to "equal" it works, in that the contrasts are set
>>> up and I can > see the results. (Works regardless of Factor2 having equal
>>> or unequal > variance.) But this makes no sense because presumably the
>>> patients and > controls have unequal variance. I am using the latest
>>> version 714 > updates. Don't have any earlier updates version to try ...
>>> > 64-bit linux machine, recompiled the mex files.
>>> > > Any help is most appreciated.
>>> > > Fatima
>>> > > -----------------------------------
>>> > > > > Spatial non-sphericity (over scans) : ...estimating
>>> derivatives
>>> > Spatial non-sphericity (over scans) : ...writing resels/voxel
>>> image
>>> > Saving results :
>>> ...done
>>> > Completed : 13:38:12 -
>>> 10/05/2007
>>> > ...use the results section for assessment
>>> > > Error running job:
>>> > Improper index matrix reference.
>>> > >
>>> --
>>> 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/
>>>
>>
>>
>>
>
> --
> 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/
>
|