[color=#000000]Dear Gustavo,[/color]
[color=#000000]Sorry the description in that thread is a bit confusing (it is really describing how to "trick" CONN into running a repeated measure analysis when you have entered your data into CONN as if you had 150 single-session subjects instead of 15 subjects with 10 sessions each, which was relevant back then because at that time there was no easy way to enter into CONN session-specific structural files). These days the process is much simpler, and you can, as you suggest, simply enter a [-4.5 -3.5 -2.5 -1.5 -.5 .5 1.5 2.5 3.5 4.5] between-conditions contrast across your 10 sessions in order to perform that linear-trend analysis across sessions (and there is no need to enter any additional "within-subjects" covariate or the like). [/color]
And regarding the gPPI effects, in the second-level tab you would simply select the appropriate conditions (e.g. the 10 NF conditions -one per session-) and enter the [-4.5 ... 4.5] contrast to look only at the temporal modulation of the connectivity strength during NF only (compared to the implicit baseline).
Hope this helps
Alfonso
[i]Originally posted by Gustavo Pamplona:[/i][quote]Dear CONN users,
I'm interested in running a PPI analysis in CONN. My dataset consists of 15 subjects, each one with 10 functional sessions. There are three "tasks": NF, FB, and Bas, the latter being the baseline. Here, I want to analyze the linear trend for NF blocks over time (i.e. sessions, or 10 time-points).
So, I found in the forum (https://www.nitrc.org/forum/message.php?msg_id=11237) the suggestion of using kron(eye(15),null([-4.5 -3.5 -2.5 -1.5 -.5 .5 1.5 2.5 3.5 4.5])') as a second-level covariate. Here I am already adapting the covariate for 15 subjects and 10 sessions. I assume that this step should be in Setup --> Covariates (2nd-level) (?)
My first question is why not to use simply [-4.5 -3.5 -2.5 -1.5 -.5 .5 1.5 2.5 3.5 4.5] in Results (2nd-level) specifying the 10 conditions (in this case sessions)?
Second question: in Setup --> Covariates (2nd-level), I create the new covariate WithinSubject and try to use kron(eye(15),null([-4.5 -3.5 -2.5 -1.5 -.5 .5 1.5 2.5 3.5 4.5])') as values, but then I receive the message: Incorrect input string size (expected array size = [1x15]; entered array size = [135x150]). I guess I misunderstood what it should be done or CONN changed the way one specifies 2nd-level covariates since the time of the recommendation mentioned above?
Third question: as recommended in https://www.nitrc.org/forum/message.php?msg_id=14685, I'm defining only NF and FB blocks and leaving Bas as the implicit baseline. However, as I'm interested only in NF blocks, how to analyze the effect only in NF? Would it change the way I define the 2nd-level covariate?
Thank you very much in advance!
Gustavo Pamplona[/quote]
[color=#000000]Sorry the description in that thread is a bit confusing (it is really describing how to "trick" CONN into running a repeated measure analysis when you have entered your data into CONN as if you had 150 single-session subjects instead of 15 subjects with 10 sessions each, which was relevant back then because at that time there was no easy way to enter into CONN session-specific structural files). These days the process is much simpler, and you can, as you suggest, simply enter a [-4.5 -3.5 -2.5 -1.5 -.5 .5 1.5 2.5 3.5 4.5] between-conditions contrast across your 10 sessions in order to perform that linear-trend analysis across sessions (and there is no need to enter any additional "within-subjects" covariate or the like). [/color]
And regarding the gPPI effects, in the second-level tab you would simply select the appropriate conditions (e.g. the 10 NF conditions -one per session-) and enter the [-4.5 ... 4.5] contrast to look only at the temporal modulation of the connectivity strength during NF only (compared to the implicit baseline).
Hope this helps
Alfonso
[i]Originally posted by Gustavo Pamplona:[/i][quote]Dear CONN users,
I'm interested in running a PPI analysis in CONN. My dataset consists of 15 subjects, each one with 10 functional sessions. There are three "tasks": NF, FB, and Bas, the latter being the baseline. Here, I want to analyze the linear trend for NF blocks over time (i.e. sessions, or 10 time-points).
So, I found in the forum (https://www.nitrc.org/forum/message.php?msg_id=11237) the suggestion of using kron(eye(15),null([-4.5 -3.5 -2.5 -1.5 -.5 .5 1.5 2.5 3.5 4.5])') as a second-level covariate. Here I am already adapting the covariate for 15 subjects and 10 sessions. I assume that this step should be in Setup --> Covariates (2nd-level) (?)
My first question is why not to use simply [-4.5 -3.5 -2.5 -1.5 -.5 .5 1.5 2.5 3.5 4.5] in Results (2nd-level) specifying the 10 conditions (in this case sessions)?
Second question: in Setup --> Covariates (2nd-level), I create the new covariate WithinSubject and try to use kron(eye(15),null([-4.5 -3.5 -2.5 -1.5 -.5 .5 1.5 2.5 3.5 4.5])') as values, but then I receive the message: Incorrect input string size (expected array size = [1x15]; entered array size = [135x150]). I guess I misunderstood what it should be done or CONN changed the way one specifies 2nd-level covariates since the time of the recommendation mentioned above?
Third question: as recommended in https://www.nitrc.org/forum/message.php?msg_id=14685, I'm defining only NF and FB blocks and leaving Bas as the implicit baseline. However, as I'm interested only in NF blocks, how to analyze the effect only in NF? Would it change the way I define the 2nd-level covariate?
Thank you very much in advance!
Gustavo Pamplona[/quote]