History of AFNI updates
Suggested reading for beginners
Add yourself to the AFNI map!
|
Author: Spiegelhalder, Kai (---.ukl.uni-freiburg.de)
Date: 09-21-05 10:11
Hello,
thank you for offering help and sorry that I compel you to guess what I mean.
I try to give a few details, even though I am not sure what sort of details may help. We study electrophysiological (E-) traces and fMRI datasets by recording them alternatingly for 2.5 sec (E- data) and 2.75 sec (fMRI volume acquisition time) in six subjects. We created a regressor for fMRI datasets from the E-data comprising of one value per epoch. Firstly we had the wrong assumption that we have to shift the hemodynamic response function parameters delaytime to 0 (instead of the default 2) and, because negative values for delaytime are not allowed, risetime to 3.375 (instead of the default 4) to get a shift that positiones the start of the response function in the middle of our E-acquisition interval. We were satisfied with our results, but noticed that we have to shift the position about 80 ms because we discarded more E-data from the beginning of the E-acquisition interval than we discarded E-data from the end of that interval. So, we adjust risetime to 3.455. Now we have got the matrix inverse average error and the results were pretty different to the first ones. We used 3dDeconvolve with two input stimulus time series. The first one only alternates between 0 and 1 every 10 epochs. The second one is the E-regressor with values between -1 and 1. The error only occurs in case that we don't use lags (but that is what we want to do). fMRI datasets were brain-only masked.
best regards
Kai Spiegelhalder
|
|
|
Spiegelhalder, Kai |
09-21-05 03:08 |
|
Spiegelhalder, Kai |
09-21-05 03:25 |
|
Robert Cox |
09-21-05 09:33 |
   |
Re: matrix inverse average error again new |
|
Spiegelhalder, Kai |
09-21-05 10:11 |
