Hello,

I was hoping that someone would be able to offer suggestions regarding if/how I should attempt to remove linear trends during data preprocessing given the run structure of our study. Briefly, each run consists of 60 seconds of rest, followed by 60 seconds during which subjects hold and view an object. During the last of the runs, subjects (deprived cigarette smokers) hold a cigarette, which is why we are using this atypical run structure (the cig elicits robust craving that would “taint” anything that follows, such as another baseline epoch – or even possibly another run). Do I run the risk of removing task-related signal changes that occur during the object-holding epoch if linear trend is removed from the data? We weren’t sure how large “real” signal changes typically are relative to the artifactual changes targeted by detrending. If they are comparable, it seems like removing linear trend might be a problem. Does anyone know if this is the case and, if so, are there any alternatives that I might want to consider (e.g., bandpass filtering)?


Thank you very much

Steve Wilson
University of Pittsburgh