 |
 |
Hi Jeannette, Thanks for your help with this, it's much appreciated. In regards to comparing fixations of different durations, would it help to include control conditions with durations matched to the experimental conditions? This would allow us to test the interaction with the contrast: (long experimental duration - long control duration) > (short experimental duration - short control duration). Would subtracting out these control events help account for non-linearity in the BOLD signal? Secondly, if it turns out that we need to keep stimulus presentation fixed and we're interested in the subsequent fixation event, is it better to include both events in the model even though they're correlated or leave the stimulus event unmodeled? Thanks! Jeremy On Tuesday, December 06, 2011 5:03:57 AM, Jeanette Mumford wrote: > Hi, > > I see, your fixation is the effect of interest so I guess you would > want to jitter the duration of the stimulus. The only issue is that > you want to compare trials with very different durations (fixation) > over a very wide range of time intervals and this could be problematic > and I don't know a good solution for that. It would be worth tracking > down a few of your colleagues locally and hashing out the details of > the model and then worrying about jittering. > > Cheers, > Jeanette > > On Mon, Dec 5, 2011 at 6:14 PM, Jeremy Elman <[log in to unmask] > <mailto:[log in to unmask]>> wrote: > > Hi Jeanette, > > Thank you for the reply! We are actually more interested in what > is going on during the fixation period, so would it make sense to > model this phase and leave the stimulus events unmodeled? If so, > then maybe it's more important that the ISI of fixation conditions > (long vs. short) are varied rather than the time between stimulus > and fixation onsets? > > Our hypothesis is that not only should there be a flip in activity > within our region of interest from stimulus to fixation (low to > high), but that longer duration fixations should be considered > different types of events than short fixation events. This is > because we think processes occurring in this region are only > beginning to kick in when people have time to disengage from the > externally attended stimulus. Hopefully that makes sense. > > Thanks for your help, it's much appreciated. > Jeremy > > > On Monday, December 05, 2011 3:14:39 PM, Jeanette Mumford wrote: > > Hi, > > You would leave fixation unmodeled and then it acts as the > baseline. So if you had a single task and fixation you would > have a single regressor that models your task and the > parameter estimate corresponding to this task would be the > activation of task compared to baseline. > > In general you must have at least 1 thing left unmodeled to > serve as the baseline in your model and fixation is typically > this thing. > > Jittering fixation time would help tease apart the signal of > two different trials types (vs baseline) that occur before and > after the fixation, which isn't what you have here. > > Hope that helps, > Jeanette > > On Mon, Dec 5, 2011 at 4:53 PM, Jeremy Elman > <[log in to unmask] <mailto:[log in to unmask]> > <mailto:[log in to unmask] <mailto:[log in to unmask]>>> wrote: > > Hello, > > I have a question regarding experimental design if we are > looking > to isolate delay activity during a memory encoding > experiment. We > would like to have subjects encode a word (2.5s or so), > followed > by a variable length fixation period (.5-8s). Our question of > interest is whether a given region of the brain which tends to > have low (or even suppressed compared to baseline) levels of > activity during the stimulus encoding phase may come > on-line given > enough time during the following fixation period. We have a few > questions about the best way to examine this though as the > encoding and fixation regressors would be correlated. > > Assuming the stimulus presentation (encoding period) is of > a set > length, the fixation onset is completely correlated with the > stimulus onset. Is there any good way to isolate this fixation > activity? My impression was that orthogonalising wrt to the > stimulus regressor wouldn't be sufficient to get around > this problem. > > Does the variable duration of the fixation phase help to > disentangle these two events or are the correlated onsets still > the main issue? We could vary the stimulus encoding duration > independently, but this isn't ideal behaviorally as it would be > unclear whether the same processes are occurring for the > duration > of the encoding event at longer times. Any other ideas on > how to > approach this problem? > > Thank you in advance for your help, > Jeremy > > > -- Jeremy A. Elman Department of Psychology University of California, Berkeley Room 3210, Tolman Hall #1650 Berkeley, CA 94720-1650 510-643-5371 (phone) 510-642-5293 (fax) [log in to unmask]