help
48 Subscribers 
help > gPPI 2nd level negative contrast
Showing 1-1 of 1 posts
Jul 6, 2016 12:07 PM | Leonardo Tozzi
gPPI 2nd level negative contrast
Dear experts,
While using gPPI I have incurred in the following situation, that led me to a methodological question.
I have a task with 2 conditions. In the first level I have calculated 2 t-tests for each subject between the interaction terms as follows: Cond1>Cond2 and Cond2>Cond1.
I then wanted to test the main effects of the two contrasts across all subjects, so I brought each of them to the second level with a one-sided t-test and then evaluated the results of each with a [1] t-contrast.
If I understand correctly, this shows the regions in which the interaction term is, respectively, significant for Cond1>Cond2 and Cond2>Cond1 across all subjects. The interpretation of this would then be, respectively, along the lines of: there is greater functional coupling between the 2 regions in condition 1 compared to condition 2 and there is greater functional coupling between the 2 regions in condition 2 compared to condition 1.
Here is my question. If in addition to running the [1] t-contrast at the second level I ran a [-1] contrast, would this be equivalent of testing across all subjects the main effect of the "inverted" first level contrast?
If so, would these two procedures then be equivalent and if not, what is the difference and the preferred one:
1) To take Cond1>Cond2 and Cond2>Cond1 to two second level one-group t-test models and run in each of them a [1] t-contrast;
2) To only take the Cond1>Cond2 to the second level and use a [1] and a [-1] t-contrast?
From my understanding, the "con" values that are produced by the first level contrasts also include negative values, that must then be explicitly tested for at the second level, since the t-tests are one-tailed. So basically, the results and interpretation of the two approaches should be the same because the "con" images are just inverted by sign. What I wonder is if the results (or their interpretation) change if the contrast comes from a negative t-test at the second level or from the opposite contrast in the first level.
Thank you very much for your help,
Leonardo Tozzi
While using gPPI I have incurred in the following situation, that led me to a methodological question.
I have a task with 2 conditions. In the first level I have calculated 2 t-tests for each subject between the interaction terms as follows: Cond1>Cond2 and Cond2>Cond1.
I then wanted to test the main effects of the two contrasts across all subjects, so I brought each of them to the second level with a one-sided t-test and then evaluated the results of each with a [1] t-contrast.
If I understand correctly, this shows the regions in which the interaction term is, respectively, significant for Cond1>Cond2 and Cond2>Cond1 across all subjects. The interpretation of this would then be, respectively, along the lines of: there is greater functional coupling between the 2 regions in condition 1 compared to condition 2 and there is greater functional coupling between the 2 regions in condition 2 compared to condition 1.
Here is my question. If in addition to running the [1] t-contrast at the second level I ran a [-1] contrast, would this be equivalent of testing across all subjects the main effect of the "inverted" first level contrast?
If so, would these two procedures then be equivalent and if not, what is the difference and the preferred one:
1) To take Cond1>Cond2 and Cond2>Cond1 to two second level one-group t-test models and run in each of them a [1] t-contrast;
2) To only take the Cond1>Cond2 to the second level and use a [1] and a [-1] t-contrast?
From my understanding, the "con" values that are produced by the first level contrasts also include negative values, that must then be explicitly tested for at the second level, since the t-tests are one-tailed. So basically, the results and interpretation of the two approaches should be the same because the "con" images are just inverted by sign. What I wonder is if the results (or their interpretation) change if the contrast comes from a negative t-test at the second level or from the opposite contrast in the first level.
Thank you very much for your help,
Leonardo Tozzi