I'm carrying out a DCM analysis on SCR responses during the conditioning of wins and losses associated with particular colors. The stimulus presentations (CS) are presented for an average of 3 seconds, with ITIs also for an average of 3 seconds. The unconditioned stimuli are two voices, saying 'win' or 'lose'. I'm interested in looking at whether responses in the two conditions are different from a baseline/neutral color condition, and different from each other. I've done the DCM tutorial with my own data for a shocking experiment (also with colors) and found the flexible response for anticipation of the shock and the fixed response for the delivery of the shock using the method described. However, for the wins and losses experiment I'm wondering whether the fixed response is a good description of the delivery of the voice, since it lasts longer than the primary reinforcing, instantaneous shock, and also should have more 'processing' involved since such feedback is usually integrated in learning models (in terms of updates and predictions). This made me think that perhaps the flexible response is the better one to look at for the delivery aspect (CS+), but I am unsure what time frame the flexible response covers, since in the onset file we provide just the start and end of the trial. So, to my questions :) Firstly, do you think the delivery CS+ response can in fact be captured in the usual way by the fixed response? And secondly, are the flexible estimates only reflecting the time period up until the moment of delivery/stimulus offset, or does this estimate also extend into the ITI following the trial (during which further processing of 'win' or 'lose' may occur)?
With kind regards,
Bronagh
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
sorry for the delay in responding - summer period.
During fear ocnditioning experiments, we found that for short stimuli under 4 s duration, a fixed responses at onset often captures the response quite nicely - indicating that the added degrees of freedom for a flexible response are mainly fitting noise. So you may actually try with a fixed response.
The time window you specify constrains the maximum "psychological" input into the sudomotor system. If you are worried that this may extend into the next trial, then you can make this window shorter.
The fact that the actual SCR occurs only during the next trial because of its delay should not be an issue here - the model takes care of that.
Hope this helps
Dominik
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
Thank you very much for your response. Just to confirm, are you saying that I could still use DCM and only look at the fixed response parameters, or that a standard GLM would perhaps be the best way to go?
Cheers,
Bronagh
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
we found that in fear conditioning, a DCM with fixed responses performs slightly better than a GLM - probably because it better accounts for no-interest SCR between trials.
However, for your short ITI of 3 s, I believe GLM is better suited as DCM cannot fit noise in the ITIs.
Best
Dominik
Last edit: Dominik Bach 2017-07-24
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
Dear Dominik,
I'm carrying out a DCM analysis on SCR responses during the conditioning of wins and losses associated with particular colors. The stimulus presentations (CS) are presented for an average of 3 seconds, with ITIs also for an average of 3 seconds. The unconditioned stimuli are two voices, saying 'win' or 'lose'. I'm interested in looking at whether responses in the two conditions are different from a baseline/neutral color condition, and different from each other. I've done the DCM tutorial with my own data for a shocking experiment (also with colors) and found the flexible response for anticipation of the shock and the fixed response for the delivery of the shock using the method described. However, for the wins and losses experiment I'm wondering whether the fixed response is a good description of the delivery of the voice, since it lasts longer than the primary reinforcing, instantaneous shock, and also should have more 'processing' involved since such feedback is usually integrated in learning models (in terms of updates and predictions). This made me think that perhaps the flexible response is the better one to look at for the delivery aspect (CS+), but I am unsure what time frame the flexible response covers, since in the onset file we provide just the start and end of the trial. So, to my questions :) Firstly, do you think the delivery CS+ response can in fact be captured in the usual way by the fixed response? And secondly, are the flexible estimates only reflecting the time period up until the moment of delivery/stimulus offset, or does this estimate also extend into the ITI following the trial (during which further processing of 'win' or 'lose' may occur)?
With kind regards,
Bronagh
Dear Bronagh
sorry for the delay in responding - summer period.
During fear ocnditioning experiments, we found that for short stimuli under 4 s duration, a fixed responses at onset often captures the response quite nicely - indicating that the added degrees of freedom for a flexible response are mainly fitting noise. So you may actually try with a fixed response.
The time window you specify constrains the maximum "psychological" input into the sudomotor system. If you are worried that this may extend into the next trial, then you can make this window shorter.
The fact that the actual SCR occurs only during the next trial because of its delay should not be an issue here - the model takes care of that.
Hope this helps
Dominik
Dear Dominik,
Thank you very much for your response. Just to confirm, are you saying that I could still use DCM and only look at the fixed response parameters, or that a standard GLM would perhaps be the best way to go?
Cheers,
Bronagh
Hi Bronath
we found that in fear conditioning, a DCM with fixed responses performs slightly better than a GLM - probably because it better accounts for no-interest SCR between trials.
However, for your short ITI of 3 s, I believe GLM is better suited as DCM cannot fit noise in the ITIs.
Best
Dominik
Last edit: Dominik Bach 2017-07-24