Hi all,
I posted this message a few weeks ago, but there was no reply. If anyone has any thoughts I'd be very grateful...
I am trying to get something straight in my head about design efficiency and the high-pass filter.
I know that efficiency depends on the contrast that you are interested in, and that contrasting events that are close together in time is good. But, I'm wondering if the high-pass filter places a hard limit on how far apart the instances of 1 event type in a model can be in time?
I have a design in which I have multiple versions of stimuli (male and female talkers) used to compose each of four conditions (A-D). I would like to compare each condition against rest, as well as against the other conditions. I would also like to be able to contrast the gender, although this is of lesser interest.
I am contemplating a design that alternates conditions A-D with silence, and the stimulus blocks need to be at least 15 seconds long, so I have made the baseline (implicitly modelled)15 seconds as well:
Am base Bm base Cm base Dm base
Af base Bf base Cf base Df base
Bm base Dm base Cm base Am base
etc. (the order of A to D in a block doesn't matter, so I figured I'd vary it).
This design is, I believe, fine if I model my conditions as A-D, and ignore gender, as the frequency of event types as well as any of the contrasts I'm interested in are <128 s apart. And, I guess I can also look at gender by making a new model where I ignore condition and just model everything as male and female.
However, is this possible to do in the SAME model (that is, putting in 8 separate event types spaced 4 minutes apart)? Or will the standard high-pass filter simply remove my design, regardless of the frequency of the subsequent contrasts, because the events are always >128 seconds apart (e.g., Am)
Many thanks in advance,
Jamie
|