Dear Pavol,

For your 2x2x2x2 factorial design (two subject groups, defined in Setup.Covariates.SecondLevel, 4 conditions, defined in Setup.Conditions, and 2 seeds, defined in Setup.ROIs) CONN will create the design matrices for you automatically and you only need to specify what particular contrasts you would like to perform by selecting the corresponding Between-subject contrast, Between-condition contrast, and Between-sources contrast in the Second-level results tab. For example:

Main effect of group: (average difference in connectivity between the two groups across all four conditions and two sources)
In the between-subject effects list select patients and controls (contrast [1 -1])
In the conditions list select SadFaces_NeutralInduction, NeutralFaces_NeutralInduction, SadFaces_SadInduction, NeutralFaces_SadInduction (contrast [1/4 1/4 1/4 1/4])
In the sources list select LeftAmygdala and RightAmygdala (contrast [1/2 1/2])

Main effect of faces: (average difference in connectivity between the sad and neutral faces across both induction conditions, both subject groups, and both sources)
In the between-subject effects list select patients and controls (contrast [1/2 1/2])
In the conditions list select SadFaces_NeutralInduction, NeutralFaces_NeutralInduction, SadFaces_SadInduction, NeutralFaces_SadInduction (contrast [1/2 -1/2 1/2 -1/2])
In the sources list select LeftAmygdala and RightAmygdala (contrast [1/2 1/2])

Main effect of induction: (average difference in connectivity between the sad and neutral induction conditions, across both faces conditions, both subject groups, and both sources)
In the between-subject effects list select patients and controls (contrast [1/2 1/2])
In the conditions list select SadFaces_NeutralInduction, NeutralFaces_NeutralInduction, SadFaces_SadInduction, NeutralFaces_SadInduction (contrast [-1/2 -1/2 1/2 1/2])
In the sources list select LeftAmygdala and RightAmygdala (contrast [1/2 1/2])

Main effect of laterality: (average difference in connectivity between the left and right seeds, across all four conditions, and both subject groups)
In the between-subject effects list select patients and controls (contrast [1/2 1/2])
In the conditions list select SadFaces_NeutralInduction, NeutralFaces_NeutralInduction, SadFaces_SadInduction, NeutralFaces_SadInduction (contrast [1/4 1/4 1/4 1/4])
In the sources list select LeftAmygdala and RightAmygdala (contrast [1 -1])

Group by Faces interaction: (between-group differences in Sad vs. Neutral faces connectivity differences, across all induction conditions and two sources)
In the between-subject effects list select patients and controls (contrast [1 -1])
In the conditions list select SadFaces_NeutralInduction, NeutralFaces_NeutralInduction, SadFaces_SadInduction, NeutralFaces_SadInduction (contrast [1/2 -1/2 1/2 -1/2])
In the sources list select LeftAmygdala and RightAmygdala (contrast [1/2 1/2])

Group by Induction interaction: (between-group differences in Sad vs. Neutral induction connectivity differences, across all faces conditions and two sources)
In the between-subject effects list select patients and controls (contrast [1 -1])
In the conditions list select SadFaces_NeutralInduction, NeutralFaces_NeutralInduction, SadFaces_SadInduction, NeutralFaces_SadInduction (contrast [-1/2 -1/2 1/2 1/2])
In the sources list select LeftAmygdala and RightAmygdala (contrast [1/2 1/2])

Group by laterality interaction: (between-group differences in left vs. right amygdala connectivity differences, across all four conditions and two sources)
In the between-subject effects list select patients and controls (contrast [1 -1])
In the conditions list select SadFaces_NeutralInduction, NeutralFaces_NeutralInduction, SadFaces_SadInduction, NeutralFaces_SadInduction (contrast [1/4 1/4 1/4 1/4])
In the sources list select LeftAmygdala and RightAmygdala (contrast [1 -1])

... (etc. all third-level interactions)

Group by faces by induction by laterality interaction:
In the between-subject effects list select patients and controls (contrast [1 -1])
In the conditions list select SadFaces_NeutralInduction, NeutralFaces_NeutralInduction, SadFaces_SadInduction, NeutralFaces_SadInduction (contrast [1 -1 -1 1])
In the sources list select LeftAmygdala and RightAmygdala (contrast [1 -1])

Also note that some times, particularly in the presence of interactions, rather than aggregating across all of the orthogonal conditions you want to look instead at the multivariate effect (an OR conjunction) across all other conditions. For example, when looking at the group by faces interaction, instead of looking at this interaction when averaging across both of the induction conditions, some times you want to look at this interaction for any of the induction conditions. Those analyses are defined using F- contrasts, like:

Group by Faces interaction (F-contrast): (between-group differences in Sad vs. Neutral faces connectivity differences, across any of the induction conditions and any of the two amygdala sources)
In the between-subject effects list select patients and controls (contrast [1 -1])
In the conditions list select SadFaces_NeutralInduction, NeutralFaces_NeutralInduction, SadFaces_SadInduction, NeutralFaces_SadInduction (contrast [1 -1 0 0; 0 0 1 -1])
In the sources list select LeftAmygdala and RightAmygdala (contrast [1 0; 0 1])


In general F- contrast matrices are simply defined by concatenating (separated by ";") each of the individual contrasts that you want to include in your OR conjunction. In the example above the F-contrast matrix for the 4 conditions is simply the concatenation of a [1 -1 0 0] contrast, which would look at the sad vs. neutral faces contrast for the neutral induction condition alone; and a [0 0 1 -1] contrast, which would look at the sad vs. neutral faces contrast for the sad induction condition alone. Similarly the F-contrast matrix for the sources is simply the concatenation of the [1 0] contrast, which would look at the connectivity with left amygdala alone, and the [0 1] contrast, which would look at the connectivity with right amygdala alone. Hope this helps clarify.

Best
Alfonso


Originally posted by Pavol Mikolas: