Just a hunch, but the .mat file you attached doesn't end with a
newline -- that might stop it from loading properly.
- A
On Thu, May 27, 2010 at 12:44 PM, Miguel Burgaleta
<[log in to unmask]> wrote:
>
> Oops, not so fast... Sorry to bother you again, but randomise is crashing
> and the error message is:
>
> ********************************************************************************************
> Loading Data: ****************************************
> Data loaded
> ERROR: Program failed
> An exception has been thrown
> Logic error:- detected by Newmat: Maximum or minimum of null matrix
>
>
> Exiting
> ********************************************************************************************
> I googled this message and searched through the archives without success...
> Any idea of what is wrong? I am attaching my DM for the whole sample (20
> subjects), just in case you want to take a look.
> Thanks again
> Miguel
>
>
>
> On Thu, May 27, 2010 at 12:47 PM, Miguel Burgaleta
> <[log in to unmask]> wrote:
>>
>> Great, Tom, thank you very much. I see light at the end of the tunnel...
>> Best,
>> Miguel
>>
>>
>> On Thu, May 27, 2010 at 11:36 AM, Thomas Nichols
>> <[log in to unmask]> wrote:
>>>
>>> Dear Miguel,
>>> Sorry, I forgot about the contrasts...
>>> For these' EV's
>>> Cd Ed S1 S2 S3 S4
>>> C1: -1 1 0 0 0 0
>>> C2: 1 -1 0 0 0 0
>>> C3: 1 0 0 0 0 0
>>> C4: 0 1 0 0 0 0
>>>
>>> You have the contrasts you specified...
>>>
>>> C1 pre-post increase in controls is higher than in experimental subjects
>>> C2 pre-post increase in experimental subjects is higher than in controls
>>> C3 pre-post increment for the control group, and C4 for the experimental
>>> group
>>>
>>> But note that this F-contrast
>>> 1 0 0 0
>>> and this F-conrast
>>> 1 1 0 0
>>> (to see overall group differences in pre-post effects) are totally
>>> equivalent (and so 1st one is better).
>>>
>>> And the exchangeability block file is plain text, no header (sorry,
>>> inconsistent, I know).
>>> -Tom
>>>
>>> On Thu, May 27, 2010 at 9:29 AM, Miguel Burgaleta
>>> <[log in to unmask]> wrote:
>>>>
>>>> OK Tom, thanks a lot! I assume that the contrast that I set for my wrong
>>>> DM can be applied to yours, right? And what info should I include in the
>>>> "header" of the EB file (before /matrix)?
>>>> Miguel
>>>>
>>>>>
>>>>> EB Cd Ed S1 S2 S3 S4
>>>>> 1 -1 0 1 0 0 0
>>>>> 2 -1 0 0 1 0 0
>>>>> 3 0 -1 0 0 1 0
>>>>> 4 0 -1 0 0 0 1
>>>>> 1 1 0 1 0 0 0
>>>>> 2 1 0 0 1 0 0
>>>>> 3 0 1 0 0 1 0
>>>>> 4 0 1 0 0 0 1
>>>>> Where EB is the exchangeability block file, EV Cd is the control
>>>>> difference (Time2-Time1) and Ed is the experimental difference, and S1-4 are
>>>>> the subject dummy (blocking) variables.
>>>>> (And, *yes*, you need the exchangeability block file... you always need
>>>>> it whenever you have repeated measures in a randomise analysis).
>>>>> -Tom
>>>>>>
>>>>>> My contrast would be:
>>>>>> -1 1 0 0 0 0 Interaction group x time (if pre-post changes in
>>>>>> group 2 (experimental) are higher than pre-post changes in group 1
>>>>>> (control))
>>>>>>
>>>>>> If this is correct, I recall my last question in my previous post: 3)
>>>>>> Do I have to use an exchangeability-block file along with the -e flag in
>>>>>> randomise? If yes, should I add a header like in .con file? The design.grp
>>>>>> would look like this?:
>>>>>>
>>>>>> 1
>>>>>> 1
>>>>>> 1
>>>>>> 1
>>>>>> 2
>>>>>> 2
>>>>>> 2
>>>>>> 2
>>>>>>
>>>>>> Thanks,
>>>>>> Miguel
>>>>>>
>>>>>>
>>>>>>
>>>>>>
>>>>>>
>>>>>> On Wed, May 26, 2010 at 9:28 AM, Stephen Smith <[log in to unmask]>
>>>>>> wrote:
>>>>>>>
>>>>>>> HI - no, start from the paired t-test design, and then take the
>>>>>>> pre-post EV and split this into one EV for controls and one for
>>>>>>> non-controls, padding with zeros.
>>>>>>> Cheers.
>>>>>>>
>>>>>>>
>>>>>>> On 25 May 2010, at 14:25, Miguel Burgaleta wrote:
>>>>>>>
>>>>>>> Thanks Stephen. I found out the problem after looking into the
>>>>>>> scripts. I was indeed applying the warps (as tbss_non_FA does), but I was
>>>>>>> assuming that they didn't include the affine transform... So I was passing
>>>>>>> the _to_target.mat twice (in the concatenated matrix and in the original
>>>>>>> warp).
>>>>>>> My next concern is about setting randomise design and contrast
>>>>>>> matrices properly. I posted an attempt to model this but didn't get an
>>>>>>> answer. I have found, however, a post where a similar case is presented so I
>>>>>>> should be able to adapt it. Could you please say if the following is
>>>>>>> correct? Let's suppose that I have 2 controls + 2 experimental subjects, and
>>>>>>> 2 timepoints. My DM would be:
>>>>>>>
>>>>>>> con-time1 con-time2 exp-time1 exp-time2 con1 con2 exp1 exp2
>>>>>>> 1 0 0 0 1 0 0 0
>>>>>>> 1 0 0 0 0 1 0 0
>>>>>>> 0 1 0 0 1 0 0 0
>>>>>>> 0 1 0 0 0 1 0 0
>>>>>>> 0 0 1 0 0 0 1 0
>>>>>>> 0 0 1 0 0 0 0 1
>>>>>>> 0 0 0 1 0 0 1 0
>>>>>>> 0 0 0 1 0 0 0 1
>>>>>>>
>>>>>>> I would include these contrasts (not 100% sure about their
>>>>>>> interpretation though --please see below):
>>>>>>> C1: -1 1 1 -1 0 0 0 0
>>>>>>> C2: 1 -1 -1 1 0 0 0 0
>>>>>>> C3: -1 1 0 0 0 0 0 0
>>>>>>> C4: 0 0 -1 1 0 0 0 0
>>>>>>> C1 will say where the pre-post increase in FA in controls is higher
>>>>>>> than in experimental subjects
>>>>>>> C2 will say where the pre-post increase in FA in experimental
>>>>>>> subjects is higher than in controls
>>>>>>> C3 will show the significant pre-post increment for the control
>>>>>>> group, and C4 for the experimental group
>>>>>>> I would also ad an F-test to see the overall change across groups:
>>>>>>> 1 1 0 0
>>>>>>> My specific questions are: 1) How do my design + contrast look? 2) Is
>>>>>>> my interpretation correct? 3) Do I have to use an exchangeability-block file
>>>>>>> along with the -e flag in randomise? If yes, should I add a header like in
>>>>>>> .con file? The design.grp would look like this:
>>>>>>> 1
>>>>>>> 1
>>>>>>> 1
>>>>>>> 1
>>>>>>> 2
>>>>>>> 2
>>>>>>> 2
>>>>>>> 2
>>>>>>> Thanks a million for your time
>>>>>>> Miguel
>>>>>>>
>>>>>>>
>>>>>>>
>>>>>>>>
>>>>>>>> Hi - it looks like maybe you're not using the nonlinear (warp)
>>>>>>>> transformations derived from the PRE images? Have a look at the scripts to
>>>>>>>> see how these come into this - I guess you'll need to copy those across from
>>>>>>>> PRE to POST to combine with the concatenated affine transform
>>>>>>>>
>>>>>>>> Cheers.
>>>>>>>>
>>>>>>>>
>>>>>>>> On 24 May 2010, at 16:53, Miguel Burgaleta wrote:
>>>>>>>>
>>>>>>>> Hello FSLers,
>>>>>>>> I am using TBSS to process a dataset of subjects scanned at 2
>>>>>>>> timepoints (PRE and POST). At this point I am trying to do the intra-subject
>>>>>>>> registration, but the result doesn't look good. This is what I have done so
>>>>>>>> far:
>>>>>>>> 1. Apply the full TBSS pipeline to PRE subjects, generating FA,
>>>>>>>> origdata and stats directories with their respective files. Visual
>>>>>>>> inspection of all_FA.nii.gz shows very nice non-linear registration to the
>>>>>>>> default template.
>>>>>>>> 2. Apply tbss_1_preproc to POST subjects, and FLIRT the output to
>>>>>>>> the PRE FA images (in native space):
>>>>>>>> flirt -dof 6 -ref PRE_FA.nii.gz -in POST_FA.nii.gz -out
>>>>>>>> POST_FA_to_PRE.nii.gz -omat POST_FA_to_PRE.mat
>>>>>>>> POST_FA_to_PRE.nii.gz looks nicely aligned to PRE images.
>>>>>>>> 3. Concatenate this transformation matrix with that resulting from
>>>>>>>> applying tbss_2_reg to PRE subjects (_FA_to_target.mat), and overwrite the
>>>>>>>> original matrix (after backup ;):
>>>>>>>> convert_xfm -omat PRE_FA_to_target.mat -concat PRE_FA_to_target.mat
>>>>>>>> POST_FA_to_PRE.mat
>>>>>>>> 4. Prepare my POST images to run tbss_non_FA on them. For this, I
>>>>>>>> first create a POST directory where I processed my PRE data with TBSS (where
>>>>>>>> origdata, FA and stats folders are) and copy the POST_FA.nii.gz files (after
>>>>>>>> tbss_1_preproc) to that location.
>>>>>>>> When I run tbss_non_FA POST, the result is not OK (POST images
>>>>>>>> occupy the same exact area than PRE images, but brain structures don't
>>>>>>>> match...). However, if I instead feed tbss_non_FA directly with the output
>>>>>>>> from 2 (POST FA images flirted to PRE images), then the result is gorgeous.
>>>>>>>> The problem is, POST images would 'suffer' 2 interpolations and PRE's only
>>>>>>>> one (actually, POST images look a bit blurry with the latter approach).
>>>>>>>> Any idea of what I am missing?
>>>>>>>> Thanks a lot in advance!
>>>>>>>> Miguel
>>>>>>>>
>>>>>>>>
>>>>>>>> ---------------------------------------------------------------------------
>>>>>>>> Stephen M. Smith, Professor of Biomedical Engineering
>>>>>>>> Associate Director, Oxford University FMRIB Centre
>>>>>>>>
>>>>>>>> FMRIB, JR Hospital, Headington, Oxford OX3 9DU, UK
>>>>>>>> +44 (0) 1865 222726 (fax 222717)
>>>>>>>> [log in to unmask] http://www.fmrib.ox.ac.uk/~steve
>>>>>>>>
>>>>>>>> ---------------------------------------------------------------------------
>>>>>>>>
>>>>>>>>
>>>>>>>>
>>>>>>>
>>>>>>>
>>>>>>>
>>>>>>>
>>>>>>> ---------------------------------------------------------------------------
>>>>>>> Stephen M. Smith, Professor of Biomedical Engineering
>>>>>>> Associate Director, Oxford University FMRIB Centre
>>>>>>>
>>>>>>> FMRIB, JR Hospital, Headington, Oxford OX3 9DU, UK
>>>>>>> +44 (0) 1865 222726 (fax 222717)
>>>>>>> [log in to unmask] http://www.fmrib.ox.ac.uk/~steve
>>>>>>>
>>>>>>> ---------------------------------------------------------------------------
>>>>>>>
>>>>>>>
>>>>>>>
>>>>>>
>>>>>
>>>>>
>>>>>
>>>>> --
>>>>> ____________________________________________
>>>>> Thomas Nichols, PhD
>>>>> Principal Research Fellow, Head of Neuroimaging Statistics
>>>>> Department of Statistics & Warwick Manufacturing Group
>>>>> University of Warwick
>>>>> Coventry CV4 7AL
>>>>> United Kingdom
>>>>>
>>>>> Email: [log in to unmask]
>>>>> Phone, Stats: +44 24761 51086, WMG: +44 24761 50752
>>>>> Fax: +44 24 7652 4532
>>>>>
>>>>
>>>
>>>
>>>
>>> --
>>> ____________________________________________
>>> Thomas Nichols, PhD
>>> Principal Research Fellow, Head of Neuroimaging Statistics
>>> Department of Statistics & Warwick Manufacturing Group
>>> University of Warwick
>>> Coventry CV4 7AL
>>> United Kingdom
>>>
>>> Email: [log in to unmask]
>>> Phone, Stats: +44 24761 51086, WMG: +44 24761 50752
>>> Fax: +44 24 7652 4532
>>>
>>
>
>
|