We are attempting to pry the percent signal change out of our data. We are using an event related design therefore our base line is intertwined with our signal.
Previous posts have suggested using the coefficient output from 3dDeconvolve with 3dCalc to get the % signal change. Take the base line coefficients, divide them into our experimental coefficients and multiply by 100 to get a percentage.
When we tried this we got large non sensical numbers.
We then thought that maybe we need to be looking at the raw +time signal as one would due if they had a block design. Just compare the control block to the experimental block directly. We figured that one way this might be possible is to use 3dConvolve to pry out the base line and our experimental condition. We would then use 3dCalc to find the percent signal change from our Convolved base line and Convolved experimental condition. We performed this calculation on a per time point basis, in other words, we did not average the Convovle input be for we ran 3dcalc.
This left us with a graph where there is a percent signal change where the spikes were all the same height for every time point and the same height in all the voxels we looked at.
This did not make sense to us as we would expect, at least, some small measure of variability in the percent signal change over, not only the time coarse, but spatially as well.
This led us to think that maybe we were of on the wrong track. The root of our problem was the fact that we have base line coefficients in and around (10) and condition coefficients in and around (40). The % signal change, using methods posted previously, would be something like a 400% difference. We have apriori knowledge that % signal change is in and around 1 or 2 % which is why we were comforted to se a 1.5% signal change in our 3dConvolve method, but found it disconcerting that all of the individual spikes had exactly 1.5% signal change.
Of course we may be way off base, but I feel like we are dancing around the solution. Any suggestions or insights would be appreciated.
Thanks
Jeremy