Hi Donald,
Thanks for your useful response. I tried what you said regarding adding the ITI. It did not make any different as the ITI is 0.67 s. Do you think it is possible to find out the linearity effects of the grip force and the signal using a block design? As I said before, I have 48 blocks, 24 active tasks. I randomized the target level of the squeeze such that each block has a different squeeze force values. Each block has 30 squeezes with ITI 0.67 seconds. Would it be better if I increase the ITI an keep the same design or do I have to use an event design?
Thanks
On 6 May 2013, at 23:37, "MCLAREN, Donald" <[log in to unmask]> wrote:
> See inline responses below.
>
>
> On Sat, May 4, 2013 at 1:58 PM, fMRI <[log in to unmask]> wrote:
>> Dear all,
>>
>> Is it possible to change the order of the columns in a design matrix in SPM?
>
>>>> Not easily.
>
>> Basically, I was reading this paper in the effects of using parametric modulation and if I understand it correctly they put the most important column of interest in as the first one, the parametric modulation.
>
>>>> In the GLM, the order of columns does not matter. However, the order entered may effect the creation of the GLM.
>
>>
>> I asked before but I think something wrong with my result or At least I cannot interrupt them.
>>
>> What I have is a blocked mixed design data of a single subject last for ~16 min. The experimental contains task ( squeezing a ball ) and rest. It has 48 blocks, 24 task and 24 rest. I started with rest. Each block lasts for 20 seconds. The active task contains 30 squeezes but each block has a different target level... example: 20% 40% 60% of the grip strength.
>>
>> I model it as follows:
>>
>> 1) I select specify first level and all the ACTIVE onset time for each squeeze so it was 720. The duration I put 0.
>
>>>> You have specified that this is a block design, so your duration should not be 0. It should be the ITI between squeezes. Then you will get the block effect.
>
>>
>> 2) I select the parameter modulation and added the grip force values for each stimuli after I mean corrected. I used spm12 and checked orthogonal.
>>
>> 3) I selected up to the third order polynomial regression. ( I tried all the possibility)
>
>>>> Do you have a specific hypothesis of why you'd expect the HRF to behave in a U or inverted-U shape based on grip strength? Or follow a 3rd order polynomial?
>
>>
>> 4) I did with and without model interaction but it seems to be better when I add model interaction (it removes lots of false positives ) should I use it ?
>
>>>> I would not model interactions. The assumption that you have "false positives" is bad as you don't know what is true activation.
>
>>
>> In the contrast I put four names
>> Active : [ 1] for just the different between active and rest and the gives me activation in where I expect
>>
>> Sq1 :[ 0 1] for the first order and it gives me activation in the eyes ( and occipital lope) which I do not expect.
>>
>> Sq2: [ 0 0 1] for the second order and it gives me activation the opposite motor area opposite to what I think I should get.
>>
>> I could not interrupt what I have as my target or my thought is that there should be a linear effects between the strength of the force and the signal.
>>
>> When I read this paper, they changed the order of the columns depend on what they want to see and put it first. Should I do this or at least can someone please guide me to what I am missing?!
>
>>>> How many blocks do you have? How many runs do you have? Have you looked at motion in these subjects?
>
>>
>>
>> Thanks
>>
>> AS
|