Hi,

I have some questions about analysis for some fMRI data that I have collected. This is my first imaging project, any help would be appreciated. I apoligize if this is too much/not enough information.

Here are the details:
Stimuli are naturalistic audiovisual clips which vary from 2-12 seconds in length. Stimuli generally fall into three categories, but these three categories are not entirely distinct throughout the clips. There are no hidden 'events' in the clips.
The stimuli were presented in an event-related design with variable jitter. The issue that has come up is that because the stimuli are quite variable in length, we are not sure how to best set up the 1D files so that our sampling is not biased but we're also not comprimising power. I don't know much about the guts of 3ddeconvolve.

We started off with 1-D files marking the stimuli for each category with a 1 for each TR where the movie clip was playing in order to account for the variable length:
e.g. a 4 second clip would be 0001100 with two ones denoting the two TR's where the clip was on.
We are now concerned about whether this is biasing our sampling with 3ddeconvolve, particularly for very long clips where some TRs are sampled many times, and for clips where subparts of the clip does not contain the particular feature we are interested in, but are samples as part of that condition. We have considered two options:

Option 1: Match the video clips by length, and mark each triad in the same place in a TR where each clip has the feature we are interested in. This would equivicate the amount of stimulation before and after what we have labled as an event, although this would be quite variable across stimuli.

Option 2: (has the same potential sampling problems as the option above, but seems to capture the dynamic nature of the video clips) Code each of the clips and mark the 1D files according to the presense of particular features we are interested in. Thus the first Tr from a particular clip could count in one condition and the second in another.

Any insight onto potential biases, advice as to what seems most reasonable, and so forth would be greatly appreciated.

Susan