Hi Eugene
Thanks for your response,
I've had a look at the res4d files and it is as you say a close model fit to the filtered_func_data being the residuals if i understand correctly.
Unfortunately I think this won't be what I need as i'm looking to model the CNR for each of my 4 conditions. I wanted to get a measure of SD for both the signal and noise within an ROI and I think to do that I need the individual fit for each condition (please correct me if i'm mistaken).
I have seen some examples of CNR discussion papers using the following method: 1) calculate average time series of all voxels within a contrast mask for all conditions and 2) a second activation signal calculated from outsize the ROI (still within grey matter). 3) These were subtracted from each other to form a noise estimate and 4) this noise estimate is then subtracted from each ROI voxel time series to calculate noise.
I believe I could use the res4d for this if motion is not a major concern (i will have to check as I would hope the MCFLIRT correction would negate the majority of its contribution) but worry that if activation for each condition is substantially different that this will affect my SD measure or general validity of the CNR measure.
Thanks
Sarah
|