Hi
This does not quite make sense, melodic_mix is 2592 by 75 (i.e.
194400/2592) so you should only be seeing 75 values per contrast - are
you running fsl_glm against melodic_mix or against a file of size
2592 by 77? Do you mind sending in the files?
cheers
Christian
On 11 Aug 2008, at 09:39, Cornelius Werner wrote:
> Hi Christian,
>
> which file do you refer to with "inputfilename.txt"? I've got the
> following files:
>
> -> melodic_mix = 2592 194400 1950992
> -> design.mat = 2597 15564 212526
> -> contrast.con = 18 67 762
>
> And yes, melodic_mix is a concatenated time series of 12 subjects. The
> design.mat accordingly was constructed by concatenating the respective
> motion correction parameters - therefore it seems that we got the
> rank-1 timecourse and the full model fit in addition to the 75
> components, each tested against the 6 regressors (!?).
>
> My question is, what is the order of the values in the output files:
> components 1-n first, then rank-1, then full model fit, yielding
> (n+2)*(no. of regressors) values? Or is it the other way round:
> rank-1, full model fit, then components 1-n? Is there any way to check
> this?
>
> Thanks a million,
> Cornelius
>
> On Thu, Aug 7, 2008 at 11:32 PM, Christian F. Beckmann
> <[log in to unmask]> wrote:
>> Hi Cornelius,
>>
>> I'd expect the same and can't replicate your problem, what do you
>> get from
>> wc inputfilename.txt
>>
>> Is it the case that the input file name was generated by
>> concatenating the
>> individual t**.txt files? If so, yes, there can be up to two
>> additional
>> regressors (77 in your case), the first one being the rank-1 time
>> course per
>> component and the second one the full model fit (provided you did
>> specify a
>> design.mat file in the GUI). These are the two regressors which are
>> actually
>> plotted in the web report.
>> hth
>> Christian
>>
>> On 6 Aug 2008, at 21:05, Cornelius Werner wrote:
>>
>>> Dear FSL experts,
>>>
>>> with the help of Christian Beckmann, I successfully ran fsl_glm on a
>>> MELODIC mixing matrix. The design matrix contained 6 regressors,
>>> the
>>> contrasts were defined such that each of the regressors was set to 1
>>> while all others were 0, respectively. Beforehand, MELODIC had
>>> identified 75 components in the data.
>>>
>>> Therefore, using the --out_p option, I expected a set of 6x75
>>> p-values. What I got, though, was a file with 6x77 values. As I
>>> skipped some lessons in statistics in school, I am really confused
>>> about this - where do the additional values come from? And how can I
>>> assign the p-values to the components?
>>>
>>> In a similar vein, I am interested in how a combination of the
>>> regressors would fit to the data. Therefore, I also ran the --out_f
>>> option. Again, with 77 values as output. Apart from having the same
>>> problem as above in assigning the f-values to the components, how
>>> do I
>>> determine the respective degrees of freedom (which I need to check
>>> for
>>> significance)?
>>>
>>> Any help is greatly appreciated - thanks in advance!
>>> Cornelius
>>>
>>> --
>>> Dr. med. Cornelius J. Werner
>>> GnuPG Public Key ID: 0x9D64BC2C
>>>
>>> Institute of Neuroscience and Biophysics - INB3
>>> Research Centre Juelich
>>> 52425 Juelich
>>> Germany
>>>
>>> Department of Psychiatry and Psychotherapy
>>> RWTH Aachen University
>>> Pauwelsstr. 30
>>> 52074 Aachen
>>> Germany
>>>
>>> ::: Please encrypt confidential data :::
>>
>
>
>
> --
> Dr. med. Cornelius J. Werner
> GnuPG Public Key ID: 0x9D64BC2C
>
> Institute of Neuroscience and Biophysics - INB3
> Research Centre Juelich
> 52425 Juelich
> Germany
>
> Department of Psychiatry and Psychotherapy
> RWTH Aachen University
> Pauwelsstr. 30
> 52074 Aachen
> Germany
>
> ::: Please encrypt confidential data :::
|