Hi,
We are attempting to compare two different impulse response magnitudes in an event related design.
We are using a 1D file that consists of 100000 for each stimuli. This represents a TR stimulus presentation followed by 5 TR's of cross hair seen on a visual screen in the magnet.
We want to compare when someone correctly identifies our stimulus vs when they do not. We wish to use the area under the curve to compare.
My question is concerning what glt matrix we should use.
We initially used 000000001-1(we have 4 concatinated runs) with a preconvolved input and max lag of 0.
Theoretically this should consider each sperate preconvolved IRF as a single multiplier which attempts to fit the wavered stimuls to the whole BOLD response. This multiplier for each IRF would be the areas that are compared. That is why we used only 1 and -1. Right?
Now what if I used a glt matrix of 00000000111111-1-1-1-1-1-1
and set the max lag to 5 with the preconvolved input. I realize that the output for each time point would not mean much, but would the output for the comparison also not mean much? Is this an interpretation of the manuals indication that when comparing two diff. IRF's we want to "approximate 'area under the curve' by summing the IRF parameters over all time lags.?"
So basically should we use all lags when using a glt with a preconvolved input, and if not why, or at least what would happen if we did?
I apologize for the length of my question(s). Any input would be greatly appreciated though.
Thank you so much for your time!
Jeremy