[color=#000000]Dear Wouter,[/color]
[color=#000000]That's an interesting question, and yes, you are exactly right you could do that by manually adding a series of first-level covariates (one per FIR time-delay) for every task condition, and then simply including these covariates in the '[i]confounds'[/i] list during the [i]Denoising [/i]step.[/color]
[color=#000000]If you have an event-related design, and are already entering your design information into CONN, one trick to have CONN create those first-level covariates automatically for you would be the following:[/color]
[color=#000000]1) in [i]Setup.Conditions[/i] select your task conditions and select the '[i]temporal decomposition (sliding window)[/i]' option. Then enter in the 'sliding-window onsets' field the values "0:TR:18" (without the quotes and changing TR for your actual TR value, in seconds), and enter in the 'sliding-window length' field the value 0. This will create a series of new conditions modeling the corresponding FIR components (up to 18s). [/color]
2) in [i]Setup.Basic [/i]and change there the 'continuous' acquisition type to 'sparse'
[color=#000000]3) in [i]Setup.Conditions [/i]select all your new FIR-component conditions (named something like 'task x Time1', 'task x Time2', etc.) and click on '[i]condition tools[/i]' and select there the option that reads '[i]move selected conditions to covariates list[/i]'. This will delete your newly created conditions and create instead a series of new first-level covariates modeling the desired FIR-components (without any form of hrf convolution, thanks to step (2) above). [/color]
4) (clean-up) in [i]Setup.Conditions s[/i]elect your original task conditions and simply revert the 'time-frequency decomposition' field to the original 'no decomposition' value, and in [i]Setup.Basic [/i]revert the 'acquisition type' field to its original 'continuous' value.
[color=#000000]Hope this helps[/color]
[color=#000000]Alfonso[/color]
[i]Originally posted by Wouter De Baene:[/i][quote]Dear all,
I was wondering what would be the best way to apply the FIR task regression approach described by Cole et al. (2019, see https://doi.org/10.1016/j.neuroimage.2018.12.054) in CONN. Does it suffice to add a series of regressors (one per time point) for every task condition as covariates in the setup?
Best regards,
Wouter De Baene[/quote]
[color=#000000]That's an interesting question, and yes, you are exactly right you could do that by manually adding a series of first-level covariates (one per FIR time-delay) for every task condition, and then simply including these covariates in the '[i]confounds'[/i] list during the [i]Denoising [/i]step.[/color]
[color=#000000]If you have an event-related design, and are already entering your design information into CONN, one trick to have CONN create those first-level covariates automatically for you would be the following:[/color]
[color=#000000]1) in [i]Setup.Conditions[/i] select your task conditions and select the '[i]temporal decomposition (sliding window)[/i]' option. Then enter in the 'sliding-window onsets' field the values "0:TR:18" (without the quotes and changing TR for your actual TR value, in seconds), and enter in the 'sliding-window length' field the value 0. This will create a series of new conditions modeling the corresponding FIR components (up to 18s). [/color]
2) in [i]Setup.Basic [/i]and change there the 'continuous' acquisition type to 'sparse'
[color=#000000]3) in [i]Setup.Conditions [/i]select all your new FIR-component conditions (named something like 'task x Time1', 'task x Time2', etc.) and click on '[i]condition tools[/i]' and select there the option that reads '[i]move selected conditions to covariates list[/i]'. This will delete your newly created conditions and create instead a series of new first-level covariates modeling the desired FIR-components (without any form of hrf convolution, thanks to step (2) above). [/color]
4) (clean-up) in [i]Setup.Conditions s[/i]elect your original task conditions and simply revert the 'time-frequency decomposition' field to the original 'no decomposition' value, and in [i]Setup.Basic [/i]revert the 'acquisition type' field to its original 'continuous' value.
[color=#000000]Hope this helps[/color]
[color=#000000]Alfonso[/color]
[i]Originally posted by Wouter De Baene:[/i][quote]Dear all,
I was wondering what would be the best way to apply the FIR task regression approach described by Cole et al. (2019, see https://doi.org/10.1016/j.neuroimage.2018.12.054) in CONN. Does it suffice to add a series of regressors (one per time point) for every task condition as covariates in the setup?
Best regards,
Wouter De Baene[/quote]