The problem with halving the dPSD is that you double the number of points in the FIR window and so you double the number of regressors and regression coefficients. Each regression coefficient is an average, so you are decreasing the number of observations going into that average by a factor of two, so the variance of the average increases. Remember that effiency is the amount of variance reduction you get from your design. It does not really have anything to do with how well the model actually fits the data -- the assumption is that your model (whatever it is) is sufficient. If you are eventually going to assume a shape of the HRF when you go to analyze your data, then there is no effiency penalty to going to finer dPSDs.
doug
Julia Bender wrote:
Hi, I have a question concerning optseq2.
I found that increasing the dPSD Parameter from 1/2 TR (0.92s) to 1 TR (1.84s) INCREASED the efficiency of the best schedules from about .075 to .093. I would have assumed that a more finely spaced schedule would result in better efficiency scores. Did I do (or understand) something wrong? My optseq2 command line goes as follows:
/.../.../optseq2 \ --ntp 121 \ --tr 1.84 \ --tprescan 0 \ --psdwin 0 11.04 0.92 \ ## or 0 11.04 1.84 --ev trial 5.52 20 \ --ev catch 5.52 20 \ --nsearch 100 \ --nkeep 10 \ --o test \
Thank you for your support,
Julia
Julia Bender Humboldt Universität zu Berlin Mathematisch - Naturwissenschaftliche Fakultät II Institut für Psychologie, Abt. Klinische Psychologie Unter den Linden 6 D-10099 Berlin
Freesurfer mailing list Freesurfer@nmr.mgh.harvard.edu https://mail.nmr.mgh.harvard.edu/mailman/listinfo/freesurfer