Hi jesper
Thanks for the speedy response to my question. I think I didn't explain
the format of the experiment well enough. One minute is the longest that
any stimuli (a fixation on crosshairs, studying words, responding "yes" or
"no" as to whether words have been seen before, and clicking a button in
response to words on the screen) is on the screen. Given that, should I use
a High Pass Filter of 2*60 for 120? Or would this be too short. 720 seems
quite long- I played around with filters bigger than 360, and got a lot of
what looks like the low freq. noise you talked about
thanks
karen-
-----Original Message-----
From: Jesper Andersson <[log in to unmask]>
To: Karen Anderson <[log in to unmask]>; spm
<[log in to unmask]>
Date: Wednesday, October 20, 1999 5:10 AM
Subject: Re: High Pass Filter
>Dear Karen,
>
>
>>hi all
>> I am working on an fMRI analysis in SPM, using the Medx based version,
>>and am having trouble deciding what number of seconds to use for the High
>>Pass Filter. I have 4 conditions in the study: studying and remembering
>>words, which runs for 3 minutes total (each stimuli displayed for 3
seconds,
>>and average amount of time studying word lists is one minute, average
amount
>>of time being tested on word lists is one minute, so each subject studies
a
>>word list, gets tested on it, and then studies another word list during
the
>>3 minutes); studying easy word lists of only one word each (three seconds
>>studying, three seconds being tested, which runs for one minute total);
>>fixation on crosshairs for one minute; and pressing a response button
>>whenever a word appears on the screen, which also runs for one minute.
>> I have tried various High Pass Filter values from 30 seconds (which I
>>think was too short and produced lots of "garbage") to 180 second and
then
>>360 seconds, which produced somewhat similar results, at least for the
>>memory conditions.
>> Does anyone have advice on what length High Pass Filter I should use?
>
>Let me see if I got you right. You have four "block" conditions, each with
>an SOA (time between onsets of consecutive stimulus of the same type) of 6
>minutes, one with a duration of three minutes, and three with a duration of
>one minute. Please correct me if I got this wrong.
>
>Given the above, the fundamental period of your experiment is 6 minutes,
>and your fundamental frequency hence 1/6*60 Hz. The point to remember about
>the high pass filter is that you do NOT want to remove any of your
>experimentally induced variance. Hence, you want your cut-off frequency for
>the HPF to be well below your fundamental frequency, e.g. half of that.
>Therefore I would suggest you using a 12*60=720 seconds HPF.
>
>Now, this is quite long and leaves you exposed to quite a bit of low
>frequency noise, but is really a consequence of the experimental design.
>Generally speaking, if the task so allows, it is an advantage to have a
>somewhat higher experimental frequency, say with a fundamental period in
>the order of 1-2 minutes.
>
>> thanks
>> karen
>
> Good luck Jesper
>Jesper Andersson
>Wellcome Dept. of Cognitive Neurology
>12 Queen Square
>London WC1N 3BG
>phone: 44 171 833 7484
>fax: 44 171 813 1420
>
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
|