Dear Claude,
> I have been told different methods to calculate the cut-off period for
> the high pass filter. The differences occur when the block design
> contains some complexity. I would therefore like to know how we should
> proceed in such cases like ours.
>
> In a recent experiment, our RUN had 8 blocks (7 scans each) separated
> by rest periods (3 scans each). One scan lasted 5.8 seconds. There were
> 4 types of tasks (images to remember) in this RUN and they were given 2
> times. The tasks were counterbalanced between subjects except the last
> 2 negatives blocks. Here is an example of subject 1:
>
> Rest 3 scans
> Positives 7 scans
> Rest 3 scans
> Neutrals 7 scans
> Rest 3 scans
> Symbolics 7 scans
> Rest 3 scans
> Positives 7 scans
> Rest 3 scans
> Neutrals 7 scans
> Rest 3 scans
> Symbolics 7 scans
> Rest 3 scans
> Negatives 7 scans
> Rest 3 scans
> Negatives 7 scans
>
> The negatives images are at the end because they are highly emotional
> and could contaminate other blocks if they are placed before. As you
> can see, Positives, Neutrals, Symbolics are presented twice in the same
> order.
>
> The traditional method to calculate the cut-off period would be to take
> the longest On-Off cycle, which in this case was rest (3 scans) + task
> (7 scans), multiply by 2 and multiply by the duration of one scan (5.8
> sec). It gives (3+7) * 2 * 5.8 sec = 116 seconds.
>
> Someone else indicated that in our block design, because of its nature,
> it should be (Rest + Positives + Neutrals + Symbolics) times 2 times
> 5.8 seconds. That gives (3+7+7+7) * 2 * 5.8sec = 278.4 sec. It is so
> because our longest cycle is in fact one series of 3 blocks presented
> twice (this method excludes the 2 negatives blocks at the end of the
> RUN).
>
> I am not sure if the repetition of 3 blocks could be consider a cycle
> in itself. Should the simple rule of the longest On-Off cycle be
> sufficient in this case to calculate the cut-off period?
Yes. Differences among Positives, Neutrals and Symbolics would be
expressed as time-varying modulations of the experiment's fundamental
frequnecy (block x time interactions) and will not be modeled by low
frequency confounds (which model the main effect of time). The simple
way to compute the cut-off period is to take the (2x) maximum interval
between epochs of the condition that occurs most frequnectly (this
condition can be thought of as providing a reference for drifts).
I hope this helps - Karl
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
|