if the correlation r = 1, that mean the two results are the same,
statistically.
usually, the signal change in block design is around 1%, for event
design 0.1%, about 10 times difference
in block design, you may set the duration as 10second.
Gao
On 24 April 2012 09:59, Ming-Tsung Tseng <[log in to unmask]> wrote:
> Hi all,
>
> My fMRI running is just alternating “on” (10 sec) and “off” (20 sec) paradigm. Each subject thus has a first-level t contrast. In another 2nd-level analysis, these 20 t contrasts were also used to correlate with a behavior score. Now I want to know the % signal change (%SC) in a specific ROI in all my 20 subjects.
>
> METHOD 1
> In the beginning, I extract %SC in this ROI by MarsBar individually:
> (1) (matlab directory is in the t contrast of subject 1) “design” – “set design from file” – select the contrast image of subject 1
> (2) “data” – “extract ROI data (default” – select this ROI
> (3) “Result” – “estimates results”
> (4) “Results” – “% signal change” – event 1 = 10, event 2 = 20. Then I got the %SC (value in event 1 – value in event 2)
>
> Am I right?
>
> By this method, I can finally get 20 numbers (values) representing %SC in this ROI for my group. But that means, I have to repeat these steps 20 times…..time-wasting…..
>
> METHOD 2
> Then I explored the second method:
> (1) (matlab directory is in the 2nd-level analysis file now!) “design” – “set design from file” – select the contrast image of 2nd-level SPM file.
> (2) Same as (2) above.
> (3) Same as (3) above.
> (4) “Data” – “Export data” – “Export summary timecourse for region” – “Export to text file”
>
> Happily, I opened this text file and I found 20 numbers. However……they were different from those obtained from method 1. But intriguingly, I found the values from both methods were HIGHLY CORRELATED (simple linear regression p<0.0001, r = 1, perfectly fitted to a line which passes through the origin 0,0)!
>
> So, which method represents real %SC (or is better) and which method should I use? What’s the relationship between values obtained from both methods? If method 2 is now allowed, any TIME-SAVING way to get the value (instead of method 1)?
>
> Thanks a lot!
>
> Carlos
|