I tried both methods as explained in "Sample Data Analysis with 3dDeconvolve". For all of my runs except one, the second method works better (where IRESP is deconvolved out given the stimulus). However, this method gives almost no activation in this one run (even though the IRESP functions peak nicely at index=2 ), whereas the first method using waver gives a spectacular activation map. The activation can also be clearly seen in the raw time series. How can there be such a discrepancy? The map is slightly better if I view it under Full F-stats instead of picking a time lag, but I'm not happy about doing that.
Please advise,
thank you,
cathy