Dear SPM users,
The aim of my experiment is to analyse the prestimulus activity, therefore, I used a finite impulse response approach.
I know that the usual bin size is 1TR but in my first the version of the batch I made a mistake and I used bin size of 1 second (ie length = 22, order =22, onset = stimulus onset – 5 ; everything is specified in seconds)... I had a look on the results and I found a nice and specific prestimulus activation (5th bin) exactly where it was predicted - though at a liberal threshold (p<0.01 uncorrected) - at the group level.
Since my TR is 1.022s (Data were acquired with a multiband sequence), I was quite confident it shouldn’t change a lot the results: I re-ran the batch with a bin size of 1.022s (ie length = 22*TR, order =22, onset = stimulus onset – 5 * TR). I was wrong : there was no more activation (even at a very liberal threshold) and when I extracted the value of the contrast in my previously activated region, there was a mean difference of 40% between the first and the second version of the analysis (across subjects).
I know that the usual bin size is 1TR but is there any reason to expect (and to explain) such a big change in the results with such a small difference in the bin size (0.022s) ?
Thank you for your help!
Fabien
|