Hi Stephanie,
Stephanie Burnett wrote:
> Hi Will,
>
> You may remember that you gave me some advice on looking at PPIs in
> fMRI data some weeks ago. This advice was very helpful and I have
> looked at PPIs in single subjects, across two subject groups and in a
> between-groups comparison. I now have a further question.
>
> I have been told to expect that activity within my seed region (a
> volume in medial PFC) will be significantly correlated with itself.
Yes, but this will only show up for one of the regressors.
To recap: you have (at least) 3 regressors in your PPI design matrix.
The first is the physiological variable (ie. the time series in the
source region), the second is the psychological variable (ie. the
experimental variable that putatively changes the correlation between
regions) and the third is the interaction regressor.
If you do a t-contrast over the 1st variable you should see your source
region lighting up.
Best,
Will.
> i.e. when I look at the T-map for PPIs within a certain contrast
> (condition A minus condition B) I should see medial PFC activity. I
> don't.
>
> Do you think this is (a) because I've made a mistake in the analysis,
> (b) because I smoothed single subject T-maps before taking them to the
> 2nd level to look at activity across the whole group, or (c) because
> you'd expect just as much auto-correlation in each condition which
> makes up the contrast?
>
> Thanks in advance for your help,
>
> Stephanie Burnett
>
> On 5/6/08, Will Penny <[log in to unmask]> wrote:
>> Hi Stephanie,
>>
>> Stephanie Burnett wrote:
>>> Hi Will,
>>>
>>> Thank you.
>>>
>>>
>>>> You can then take these con images to the second level in the usual way
>> (ie.
>>>> do another SPM analysis, a one sample t-test, with these con images as
>> the
>>>> data).
>>>>
>>> For group (second level analysis), would you advise
>>>
>>> (A) defining a VOI separately in each subject (centred on the peak
>>> voxel within a brain region e.g. V2 for the psychological contrast of
>>> interest), and then taking all these (partially overlapping) volumes
>>> up to the 2nd level for group analysis, or
>>>
>>>
>> I would go with this option - option A.
>>
>> You can always smooth the contrast images further, to accomodate
>> between-subject differences in functional anatomy and improve the overlap.
>>
>> Best,
>>
>> Will.
>>
>>
>>> (B) defining the same VOI in each subject, centred on the peak group
>>> activation? My problem with (B) is that this will give quite
>>> weak/noisy activations in some subjects, as the brain region I'm
>>> taking my VOI from is quite big, so single subject activations don't
>>> always overlap.
>>>
>>> Many thanks for your very helpful guidance so far,
>>>
>>> Stephanie
>>>
>>>
>>> On 5/2/08, Will Penny <[log in to unmask]> wrote:
>>>
>>>> Stephanie Burnett wrote:
>>>>
>>>>> Hi again,
>>>>>
>>>>> Thanks very much for your help so far. I have another question.
>>>>>
>>>>> Is there is a way to collapse PPIs across sessions? Specifically, two
>>>>> sessions from the same subject or sessions from multiple subjects.
>>>>>
>>>>>
>>>>>
>>>> For multiple sessions from 1 subject you can do a single SPM analysis
>>>> with 3 sessions. Each of these will have its own PPI.ppi, PPI.P and
>> PPI.Y
>>>> variable from each of the 3 sessions.
>>>>
>>>> You can then test for the overall PPI effect across sessions using a
>>>> contrast like
>>>>
>>>> 0 0 1 0 0 1 0 0 1 0
>>>>
>>>> which aggregates the PPI effect over sessions. This test will also
>>>> produce a contrast image. One for each subject.
>>>>
>>>> You can then take these con images to the second level in the usual way
>> (ie.
>>>> do another SPM analysis, a one sample t-test, with these con images as
>> the
>>>> data).
>>>>
>>>> Best,
>>>>
>>>> Will.
>>>>
>>>>
>>>>
>>>>
>>>>> Best wishes,
>>>>>
>>>>> Stephanie
>>>>>
>>>>>
>>>>>
>>>>> On 5/2/08, Stephanie Burnett <[log in to unmask]> wrote:
>>>>>
>>>>>
>>>>>> Dear Will,
>>>>>>
>>>>>> Brilliant. I took the Y variable from the original VOI file and the
>>>>>> problem is now solved.
>>>>>>
>>>>>> Thanks,
>>>>>>
>>>>>> Stephanie
>>>>>>
>>>>>>
>>>>>> On 5/2/08, Will Penny <[log in to unmask]> wrote:
>>>>>>
>>>>>>
>>>>>>> Hi,
>>>>>>>
>>>>>>> PPI.ppi is OK
>>>>>>> PPI.P is OK
>>>>>>>
>>>>>>> but PPI.Y is not. Its entries are just 'NaN', hence the
>>>>>>> completely white (blank) column for your VOI.
>>>>>>>
>>>>>>> So I would check the steps used to create this.
>>>>>>>
>>>>>>> If there's no way round it then you could take the Y variable
>>>>>>> from the original VOI file.
>>>>>>>
>>>>>>> Best,
>>>>>>>
>>>>>>> Will.
>>>>>>>
>>>>>>>
>>>>>>> Stephanie Burnett wrote:
>>>>>>>
>>>>>>>
>>>>>>>> Dear Will,
>>>>>>>>
>>>>>>>> Thanks very much for getting back to me. The lower case p in my
>>>>>>>>
>>>> email
>>>>
>>>>>>>> was a typo - what I put into SPM was indeed a capital P.
>>>>>>>>
>>>>>>>> The PPI file is attached.
>>>>>>>>
>>>>>>>> Best wishes,
>>>>>>>>
>>>>>>>> Stephanie
>>>>>>>>
>>>>>>>>
>>>>>>>>
>>>>>>>> On 5/2/08, Will Penny <[log in to unmask]> wrote:
>>>>>>>>
>>>>>>>>
>>>>>>>>
>>>>>>>>> Stephanie Burnett wrote:
>>>>>>>>>
>>>>>>>>>
>>>>>>>>>
>>>>>>>>>> Dear Will,
>>>>>>>>>>
>>>>>>>>>> I'm working with Sarah-Jayne Blakemore, who emailed you
>>>>>>>>>>
>>>> yesterday
>>>>
>>>>>>>>>> about some difficulties we're having with PPIs. We're very
>>>>>>>>>>
>>>> grateful
>>>>
>>>>>>>>>> for your help.
>>>>>>>>>>
>>>>>>>>>> I used the link you sent
>>>>>>>>>>
>>>>>>>>>>
>>>>>>>>>>
>>>>>>>>>>
>> ftp://ftp.fil.ion.ucl.ac.uk/spm/data/attention/README_GLM_PPI.txt
>>>>>>>>>> ... to try to do a PPI on my data. It all went fine until I
>> got
>>>> to the
>>>>
>>>>>>>>>> 'fMRI - specify data' part, when an error message appeared
>> (I've
>>>>>>>>>> attached it as a .mat file in case that helps).
>>>>>>>>>>
>>>>>>>>>> Another thing (perhaps related) that looks not quite right
>> is
>>>> that the
>>>>
>>>>>>>>>> design matrix came up with a completely white (blank) column
>> for
>>>> my
>>>>
>>>>>>>>>> VOI, and the orthogonality looked strange (the VOI's cell
>> was
>>>>>>>>>> missing).
>>>>>>>>>>
>>>>>>>>>> I'm also not sure that my 'PPI_regionXcontrast.mat' file is
>>>>>>>>>>
>>>> correct.
>>>>
>>>>>>>>>> when i load it into matlab, there is no 'p' that i can see.
>> so
>>>> when i
>>>>
>>>>>>>>>> load the user-specified regressors (in the 'fMRI - specify
>>>>>>>>>>
>>>> design'
>>>>
>>>>>>>>>> stage), how does spm know what [PPI.p] is?
>>>>>>>>>>
>>>>>>>>>>
>>>>>>>>>>
>>>>>>>>>>
>>>>>>>>>>
>>>>>>>>> I think it should be captial P. See eg.
>>>>>>>>>
>>>>>>>>> 13. [360] regressor 1 [PPI.ppi]
>>>>>>>>> 14. [360] regressor 2 [PPI.P]
>>>>>>>>> 15. [360] regressor 2 [PPI.Y]
>>>>>>>>>
>>>>>>>>> If it still does'nt work can you email me this file
>>>>>>>>> ('PPI_regionXcontrast.mat') ?
>>>>>>>>>
>>>>>>>>> I'll have a look,
>>>>>>>>>
>>>>>>>>> W.
>>>>>>>>>
>>>>>>>>>
>>>>>>>>>
>>>>>>>>>
>>>>>>>>>> I would be very grateful for your help
>>>>>>>>>> best wishes
>>>>>>>>>> Stephanie
>>>>>>>>>>
>>>>>>>>>> --
>>>>>>>>>> Ms. Stephanie Burnett, PhD student
>>>>>>>>>> UCL Institute of Cognitive Neuroscience
>>>>>>>>>> 17 Queen Square
>>>>>>>>>> London WC1N 3AR, UK
>>>>>>>>>> t: 00 44 20 7679 1126
>>>>>>>>>>
>>>>>>>>>>
>>>>>>>>>>
>>>>>>>>>>
>> http://www.icn.ucl.ac.uk/Research-Groups/Developmental-Group/index.php
>>>>>>>>>>
>>>>>>>>> --
>>>>>>>>> William D. Penny
>>>>>>>>> Wellcome Trust Centre for Neuroimaging
>>>>>>>>> University College London
>>>>>>>>> 12 Queen Square
>>>>>>>>> London WC1N 3BG
>>>>>>>>>
>>>>>>>>> Tel: 020 7833 7475
>>>>>>>>> FAX: 020 7813 1420
>>>>>>>>> Email: [log in to unmask]
>>>>>>>>> URL: http://www.fil.ion.ucl.ac.uk/~wpenny/
>>>>>>>>>
>>>>>>>>>
>>>>>>>>>
>>>>>>>>>
>>>>>>>>>
>>>>>>>>
>>>>>>>>
>>>>>>> --
>>>>>>> William D. Penny
>>>>>>> Wellcome Trust Centre for Neuroimaging
>>>>>>> University College London
>>>>>>> 12 Queen Square
>>>>>>> London WC1N 3BG
>>>>>>>
>>>>>>> Tel: 020 7833 7475
>>>>>>> FAX: 020 7813 1420
>>>>>>> Email: [log in to unmask]
>>>>>>> URL: http://www.fil.ion.ucl.ac.uk/~wpenny/
>>>>>>>
>>>>>>>
>>>>>>>
>>>>>>>
>>>>>> --
>>>>>> Ms. Stephanie Burnett, PhD student
>>>>>> UCL Institute of Cognitive Neuroscience
>>>>>> 17 Queen Square
>>>>>> London WC1N 3AR, UK
>>>>>> t: 00 44 20 7679 1126
>>>>>>
>>>>>>
>> http://www.icn.ucl.ac.uk/Research-Groups/Developmental-Group/index.php
>>>>>>
>>>>>
>>>>>
>>>> --
>>>> William D. Penny
>>>> Wellcome Trust Centre for Neuroimaging
>>>> University College London
>>>> 12 Queen Square
>>>> London WC1N 3BG
>>>>
>>>> Tel: 020 7833 7475
>>>> FAX: 020 7813 1420
>>>> Email: [log in to unmask]
>>>> URL: http://www.fil.ion.ucl.ac.uk/~wpenny/
>>>>
>>>>
>>>>
>>>
>>>
>> --
>> William D. Penny
>> Wellcome Trust Centre for Neuroimaging
>> University College London
>> 12 Queen Square
>> London WC1N 3BG
>>
>> Tel: 020 7833 7475
>> FAX: 020 7813 1420
>> Email: [log in to unmask]
>> URL: http://www.fil.ion.ucl.ac.uk/~wpenny/
>>
>>
>
>
--
William D. Penny
Wellcome Trust Centre for Neuroimaging
University College London
12 Queen Square
London WC1N 3BG
Tel: 020 7833 7475
FAX: 020 7813 1420
Email: [log in to unmask]
URL: http://www.fil.ion.ucl.ac.uk/~wpenny/
|