Hi Hallvard -
It might be a concern if you were generating contrasts comparing some events that were relatively brief, say, 6 seconds, to events that were on average considerably longer, e.g. > 15 or 20 seconds, as a lower amplitude for the brief events might simply reflect the fact that the hemodynamic response is not reaching its peak. You also want to be careful if the two phases always occur in the same order - in some regions there may be some consistent temporal effect that is unrelated to the phases.
You might use a second-level regressor to determine whether duration affects the amplitude of responses to a given phase, which may help you determine whether to trust your contrasts between phases when they have different average durations? Stretching your time series is definitely not a good idea.
Eugene
2009/6/5 Hallvard Røe
<[log in to unmask]>
The differences are in the duration of the trials, because the time the participants did use depended on strategy used during navigation
and/or how well they knew the environment. I`m not comparing groups. But I`m comparing different phases during spatial navigation. The total navigation time varies, as well as the time of the different phases during navigation. The total time of each block was maximally 40 sec. For the two phases that we are most interested in comparing, it would in average be so that if the participant used a long time to complete one of these phases, then the other phase would also be long. However, in the other comparison we did, then the relationship was oppsite, meaning that if one of the phases was long then the other one would be shorter.
Hallvard
-------------------------------------------------------------------------------
Siterer Eugene Duff <
[log in to unmask]>:
Hi Hallvard -
Are the differences in the total scan time or the duration of the trials?
What is your comparison of interest (e.g. are you comparing groups)?
Some variation in, say, the number of trials across subjects does not
usually have to be explicitly accounted for, as the variation in the
reliability of pes from different sessions is taken into account in the
standard statistics.
Eugene
2009/6/5 Hallvard Røe Evensmoen <[log in to unmask]>
I have completed an fMRI experiment, in which it was a systematic
difference
between the amount of time used and the subjects. How to take this into
account when running the analysis?
- Add an average time regressor at third level
- Multiply the time periods with a factor, so that the length of the
time-periods for those was fast becomes more similar to the length of the
time periods for those that used most time
.....................
Anyone got any suggestions?
Best Regards
Hallvard Røe Evensmoen
--
Eugene Duff
FMRIB Centre,
University of Oxford
John Radcliffe Hospital, Headington OX3 9DU Oxford UK
Ph: +44 (0) 1865 222 739 Fax: +44 (0) 1865 222 717
--
--
Eugene Duff
FMRIB Centre,
University of Oxford
John Radcliffe Hospital, Headington OX3 9DU Oxford UK
Ph: +44 (0) 1865 222 739 Fax: +44 (0) 1865 222 717
--
