On 11/12/05, [log in to unmask] <[log in to unmask] > wrote:

hi,

I am also interested in discussion about this. The high-pass filter is to remove the low frequency confounds in the BOLD time series signal, and there is autoregressive (AR) model or temporal smoothing available to remove the autocorrelation. But what is the purpose of normally used low-pass filter derived from a typical HRF? Is the low-pass filter also used for correction the autocorrelation? Can we use all of them (high-pass, low-pass filter, AR) in the GLM modeling for one study?

And, if i have some real long epoch design, for example, 12 min scanning, including 3min rest -6 min task - 3min rest. How can i determine the proper filter, esp the high-pass filter? Should i calculate the high-pass cut-off as 2x(3+6) = 18 min = 1080 s. It seems too much longer than the normal value automatically suggested by SPM

Any advices are highly appreciated.

Hengyi Rao

Univ of Pennsylvania

----------------------------

Hi SPMers:

I am also very interesting in this topic, Can someone give us some details about it ? As you konw, I am not familar with the SPM routine.

Thanks

Leina

On 11/9/05, Amit Anand <[log in to unmask]> wrote:
Can some one let us know what are the procedures in SPM to filter (at a
specified value) and detrend raw data before processing it.

Thanks,

Amit Anand