I am new to PSPM and modelling the EDA data, so I have a few fundamental questions regarding the analyses of my experiment. We are presenting videos of faces changing from neutral to a fearful or happy face. The videos are about 3 seconds long (changes detection occurs at about 1,5 sec after stimulus onset) followed by a ISI of 4 seconds before the next video is presented. First, I wanted to set up a GLM in response to the stimulus onset. Based off what I was observing at the single subject level, I saw that there was a very robust negative deflection and then a later positive deflection. By plotting the raw data, I noticed that the SCR was starting before the trial onset, maybe indicating an anticipation phenomenon occurring before the neutral face occurs and at the time when it is changing to an emotional expression. We then observed an additional little peak 4-5 seconds after the stimulus onset, presumably a reaction to the emotional expression. We were wondering if it would be possible to model both of these SCRs and what would be the best way since they are potentially overlapping and the ISI is not that long. Based on the posts in the forum I would now try to run the DCM once with event windows before the trial presentation (maybe 1-2 sec before and about one after the onset of the video, because the change detection occurs about 1,5 sec after stimulus onset) and defining a time window for a response after that. Do you think this would be a possible way for us to analyse the data?
Thank you in advance!
Marleen
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
sorry for the late reply. Separating SCR in short succession and determining their amplitude is a great application of PsPM. However, what you need (as with any peak-scoring analysis) is a clear idea of when the SCR occur. Often this is precisely mandated by the experimental timing, but in your case, the pre-stimulus anticipation and the dynamic nature of the stimuli appear to preclude such deterministic approach, and DCM which allows to estimate the timing from the data, could be more appropriate.
It makes good sense to define time windows as you suggested. I would like to add that it is important to make sure the definition of these time windows is not related to your hypothesis (for DCM as for any peak-scoring analysis). In other words, when you select the time windows based on peri-stimulus SCR averages across all conditions, then you cannot use the ensuing amplitude estimates to "detect" responses (this would be circular), but you can use them to contrast different experimental conditions.
Finally, just to note that small experimental variation in SCR onset (up to 1 s) is often unproblematic for GLM, since the SCR are rather variable over trials and subjects anyway. For example, we have shown that SCR derived from the foot (where the peak occurs visibly later than at the hand, due to longer nerve conduction times) is as well fit with a GLM based on a hand-derived SCR as with a model that takes the later peak into account (Bach et al. 2010).
Hope this helps.
Dominik
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
Hi Dominik,
I am new to PSPM and modelling the EDA data, so I have a few fundamental questions regarding the analyses of my experiment. We are presenting videos of faces changing from neutral to a fearful or happy face. The videos are about 3 seconds long (changes detection occurs at about 1,5 sec after stimulus onset) followed by a ISI of 4 seconds before the next video is presented. First, I wanted to set up a GLM in response to the stimulus onset. Based off what I was observing at the single subject level, I saw that there was a very robust negative deflection and then a later positive deflection. By plotting the raw data, I noticed that the SCR was starting before the trial onset, maybe indicating an anticipation phenomenon occurring before the neutral face occurs and at the time when it is changing to an emotional expression. We then observed an additional little peak 4-5 seconds after the stimulus onset, presumably a reaction to the emotional expression. We were wondering if it would be possible to model both of these SCRs and what would be the best way since they are potentially overlapping and the ISI is not that long. Based on the posts in the forum I would now try to run the DCM once with event windows before the trial presentation (maybe 1-2 sec before and about one after the onset of the video, because the change detection occurs about 1,5 sec after stimulus onset) and defining a time window for a response after that. Do you think this would be a possible way for us to analyse the data?
Thank you in advance!
Marleen
Hi Marleen
sorry for the late reply. Separating SCR in short succession and determining their amplitude is a great application of PsPM. However, what you need (as with any peak-scoring analysis) is a clear idea of when the SCR occur. Often this is precisely mandated by the experimental timing, but in your case, the pre-stimulus anticipation and the dynamic nature of the stimuli appear to preclude such deterministic approach, and DCM which allows to estimate the timing from the data, could be more appropriate.
It makes good sense to define time windows as you suggested. I would like to add that it is important to make sure the definition of these time windows is not related to your hypothesis (for DCM as for any peak-scoring analysis). In other words, when you select the time windows based on peri-stimulus SCR averages across all conditions, then you cannot use the ensuing amplitude estimates to "detect" responses (this would be circular), but you can use them to contrast different experimental conditions.
Finally, just to note that small experimental variation in SCR onset (up to 1 s) is often unproblematic for GLM, since the SCR are rather variable over trials and subjects anyway. For example, we have shown that SCR derived from the foot (where the peak occurs visibly later than at the hand, due to longer nerve conduction times) is as well fit with a GLM based on a hand-derived SCR as with a model that takes the later peak into account (Bach et al. 2010).
Hope this helps.
Dominik