Hi,
Thank you very much for your reply.
First thing I am testing is where correlation between FA values and behavior
is larger (smaller) in patient and control group.
My design is :
Group EV1 EV2
1 2 0
1 3 0
1 -5 0
2 0 -1
2 0 -2
2 0 -3
C1 1 -1
C2 -1 1
Where EV1 and EV2 are demeaned behavioral scores, group 1 is controls, group
2 patients (my real design includes of course more participants). I included
no F-tests.
My command line would be:
randomise -i all_FA_skeletonised.nii -o output -m mask1.nii -d design.mat -t
design.con -c 1.5 -D -V
Should I include 2 further Evs, specyfying group membership?
EV3 EV4
1 0
1 0
1 0
0 1
0 1
0 1
Thank you!
Cheers,
Aga
On 3/29/09 7:59 AM, "Steve Smith" <[log in to unmask]> wrote:
> Hi - this is now the "-e" option - see the randomise manual for more
> information on this. For most simple designs you don't need this -
> we'd need to know more about your model to be sure though.
>
> Cheers, Steve.
>
>
> On 28 Mar 2009, at 15:17, Agnieszka Burzynska wrote:
>
>>
>> Dear all
>> I found this post below in the archives and wonder whether I need to
>> use the g flag for a desing similar to the one below. I dont find
>> the -g option in the current version of randomise and therefore I am
>> not sure what to do.
>>
>> My second question is: how much it matters to specify a group in the
>> first ³group² column in the Glm-GUI? Is it just for userıs own
>> information? If I left all inputs with ³1² but I specify groups in
>> the Evs( for EV only one subgroup has non-zero values, etc) , does
>> this first column still matter?
>>
>> Thank you,
>> Aga
>>
>>
>>>>> On 5 Mar 2008, at 16:55, :
>>>>>
>>>>>>> Hi Steve,
>>>>>>>
>>>>>>> Firstly I am not very familiar with the mathematics of design
>> matrix
>>>>>>> and contrast matrix. I am trying to get myself familiarized
>> with it,
>>>>>>> so kindly excuse me if any of my questions sound silly.
>>>>>>>
>>>>>>> I have two groups
>>>>>>>
>>>>>>> No of Controls: 13
>>>>>>> No of Patients: 12
>>>>>>> Behavioral data: Type1 (for both patients and controls)
>>>>>>> Behavioral data: Type 2 (for both patients and controls)
>>>>>>>
>>>>>>> I would like to set up the randomise option in TBSS to do
>>>>>>> correlation between the FA values and a behavioral measure
>> (Actually
>>>>>>> both - separately). There are two types of tests that I would
>> like
>>>>>>> to carry out.
>>>>>>>
>>>>>>> 1. Test whether the FA of all the subjects in both the
>> groups
>>>>>>> correlate with a behavioral measure.
>>>>>>> 2. Contrast the behavioral correlation between the two
>> groups.
>>>>>>>
>>>>>>> From some of the previous posts I initially assumed that I
>> needed
>>>>>>> two design matrices and one contrast matrix for both the
>> tests. But
>>>>>>> in the latest version of randomise that I downloaded it said
>> that
>>>>>>> confound regressors can be given in the main design matrix
>> itself.
>>>>>>> So I am assuming that we need only one design matrix and one
>>>>>>> contrast matrix for both of the above tests that I had
>> mentioned
>>>>>>> above.
>>>>>>>
>>>>>>> If the above assumptions I have made are right then I will
>> move on
>>>>>>> with the design and contrast matrices that I have for both of
>> the
>>>>>>> above tests.
>>>>>>>
>>>>>>> Test 2 - Contrast the behavioral correlation between the two
>> groups.
>>>>>>>
>>>>>>> So in this design matrix I had 2 groups and 4 EVs. In the
>> groups
>>>>>>> column I gave the value for first 13 subjects as 1 and for
>> the rest
>>>>>>> of the 12 subjects to be 2. The matrix given below is the
>> design
>>>>>>> matrix that I used for this testing. (b1-b25 is the behavioral
>>>>>>> values for each of the 25 subjects.)
>>>>>>>
>>>>>>> EV1 EV2 EV3 EV4
>>>>>>> b1 0 1 0
>>>>>>> b2 0 1 0
>>>>>>> b3 0 1 0
>>>>>>> b4 0 1 0
>>>>>>> b5 0 1 0
>>>>>>> b6 0 1 0
>>>>>>> b7 0 1 0
>>>>>>> b8 0 1 0
>>>>>>> b9 0 1 0
>>>>>>> b10 0 1 0
>>>>>>> b11 0 1 0
>>>>>>> b12 0 1 0
>>>>>>> b13 0 1 0
>>>>>>> 0 b14 0 1
>>>>>>> 0 b15 0 1
>>>>>>> 0 b16 0 1
>>>>>>> 0 b17 0 1
>>>>>>> 0 b18 0 1
>>>>>>> 0 b19 0 1
>>>>>>> 0 b20 0 1
>>>>>>> 0 b21 0 1
>>>>>>> 0 b22 0 1
>>>>>>> 0 b23 0 1
>>>>>>> 0 b24 0 1
>>>>>>> 0 b25 0 1
>>>>>>>
>>>>>>> From some of the past archives I found that the matrix that
>> contains
>>>>>>> the behavioral measures is the main matrix(used with -d) and
>> the
>>>>>>> matrix with the group membership information should be taken
>> as the
>>>>>>> confound matrix (used with -x).
>>>>>
>>>>> No, the group membership, which controls the exchangeability
>> blocks,
>>>>> should be passed in using the -g flag. There is no longer any
>> need for
>>>>> 'confound' matrices.
>>
>
>
> ---------------------------------------------------------------------------
> 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
> ---------------------------------------------------------------------------
|