[color=#000000]Hi Jenna,[/color]
[color=#000000]Regarding (1) in the [i]first-level analysis [/i]tab, switch the plot title that reads "[i]source timeseries[/i]" to "[i]first-level analysis design matrix[/i]" in order to display the analysis design matrix[/color]
[color=#000000]Regarding (2), you are exactly right on all accounts: gPPI is an inherently multivariate analysis, but the "regression (multivariate)" vs. "regression (univariate)" selection in CONN only refers to how multiple seeds/sources will be treated ("regression -multivariate" when you want all seeds entered jointly as multiple "physiological terms" in a single gPPI analysis, vs "regression -univariate" when you want each seed to be entered as a single/unique "physiological term" into a separate gPPI analysis -this latter case is the most common scenario in standard gPPI analyses-)[/color]
[color=#000000]Best[/color]
[color=#000000]Alfonso[/color]
[i]Originally posted by Jenna Adams:[/i][quote]Hi all,
I am performing a gPPI analysis with four experimental conditions in an event related design, with fixation crosses between stimuli as the baseline (not explicitly modeled as a condition in the set-up). I have a couple of questions about the first-level design:
1) Is it possible to see the design matrix of the first-level gPPI to get a better sense of what exactly is being modeled? I've heard other programs that use gPPI show it at this stage.
2) In the first-level specification, what is the difference between selecting "regression (bivariate)" and "regression (multivariate)"? The CONN manual said to choose one of the regression models, and from my understanding gPPIs are inherently multivariate models. What would be the reason to pick bivariate here? Or is this choice related to having more than one seed ROI included in the same gPPI model (ex. pick multivariate if you have multiple seeds)?
Thank you in advance for the help!
Jenna[/quote]
[color=#000000]Regarding (1) in the [i]first-level analysis [/i]tab, switch the plot title that reads "[i]source timeseries[/i]" to "[i]first-level analysis design matrix[/i]" in order to display the analysis design matrix[/color]
[color=#000000]Regarding (2), you are exactly right on all accounts: gPPI is an inherently multivariate analysis, but the "regression (multivariate)" vs. "regression (univariate)" selection in CONN only refers to how multiple seeds/sources will be treated ("regression -multivariate" when you want all seeds entered jointly as multiple "physiological terms" in a single gPPI analysis, vs "regression -univariate" when you want each seed to be entered as a single/unique "physiological term" into a separate gPPI analysis -this latter case is the most common scenario in standard gPPI analyses-)[/color]
[color=#000000]Best[/color]
[color=#000000]Alfonso[/color]
[i]Originally posted by Jenna Adams:[/i][quote]Hi all,
I am performing a gPPI analysis with four experimental conditions in an event related design, with fixation crosses between stimuli as the baseline (not explicitly modeled as a condition in the set-up). I have a couple of questions about the first-level design:
1) Is it possible to see the design matrix of the first-level gPPI to get a better sense of what exactly is being modeled? I've heard other programs that use gPPI show it at this stage.
2) In the first-level specification, what is the difference between selecting "regression (bivariate)" and "regression (multivariate)"? The CONN manual said to choose one of the regression models, and from my understanding gPPIs are inherently multivariate models. What would be the reason to pick bivariate here? Or is this choice related to having more than one seed ROI included in the same gPPI model (ex. pick multivariate if you have multiple seeds)?
Thank you in advance for the help!
Jenna[/quote]
