help
help > RE: How to combine r values of task and baseline
Apr 22, 2015 12:04 PM | Alfonso Nieto-Castanon - Boston University
RE: How to combine r values of task and baseline
Dear Ivan,
Yes, typically the first-level connectivity measures represent Fisher-transformed correlation coefficients, so if you select two conditions (e.g. task and baseline) and enter a between-conditions contrast [1 -1] you will be effectively looking at the within-subject difference in Fisher-transformed r coefficients between the task vs baseline conditions (simply the difference r_task - r_baseline, where r_task and r_baseline represent the Fisher-transformed correlation coefficients computed separately for each condition, respectively). In your case I would expect the typical contrast of interest to be something like the following: select 'patient' and 'control' and enter a between-subject contrast [1 -1], and select 'task1' and 'baseline' and enter a between-condition contrast [1 -1]. That will evaluate group by task interactions in your data. In practice, this means that, for each subject you are first looking at the difference in Fisher-transformed r coefficients between the task1 and baseline conditions, and then you are taking those within-subject differences across all subjects and evaluating whether they are different between the patient and control groups.
Let me know if this clarifies
Alfonso
Originally posted by Yang Yang:
Yes, typically the first-level connectivity measures represent Fisher-transformed correlation coefficients, so if you select two conditions (e.g. task and baseline) and enter a between-conditions contrast [1 -1] you will be effectively looking at the within-subject difference in Fisher-transformed r coefficients between the task vs baseline conditions (simply the difference r_task - r_baseline, where r_task and r_baseline represent the Fisher-transformed correlation coefficients computed separately for each condition, respectively). In your case I would expect the typical contrast of interest to be something like the following: select 'patient' and 'control' and enter a between-subject contrast [1 -1], and select 'task1' and 'baseline' and enter a between-condition contrast [1 -1]. That will evaluate group by task interactions in your data. In practice, this means that, for each subject you are first looking at the difference in Fisher-transformed r coefficients between the task1 and baseline conditions, and then you are taking those within-subject differences across all subjects and evaluating whether they are different between the patient and control groups.
Let me know if this clarifies
Alfonso
Originally posted by Yang Yang:
Dear Dr. Alfonso,
I have a set of task-induced fMRI data with a 2 groups (patient, control) *3 conditions (task1, task2, baseline) design. Now, I hope to examine the group difference in connectivity induced by task. Using ROI-to-ROI analysis, I extract the r value (after Z transformation) for each condition and subject. However, I just focus on how the particular task component (task minus baseline) modulate the connectivity among ROIs. Could I use difference in r value derived from r of task minus r of baseline to conduct the group analysis? or I must use a 2*3 ANOVA analysis for r value?
A related question is how 'conn' calculates the relative r value among two or more conditions? For example, if I set (1 -1) for condition contrast (task vs. baseline), is the result a effect size for r of task contrast to r of baseline? What the result represent? How to explain the results?
Thanks!
Ivan
I have a set of task-induced fMRI data with a 2 groups (patient, control) *3 conditions (task1, task2, baseline) design. Now, I hope to examine the group difference in connectivity induced by task. Using ROI-to-ROI analysis, I extract the r value (after Z transformation) for each condition and subject. However, I just focus on how the particular task component (task minus baseline) modulate the connectivity among ROIs. Could I use difference in r value derived from r of task minus r of baseline to conduct the group analysis? or I must use a 2*3 ANOVA analysis for r value?
A related question is how 'conn' calculates the relative r value among two or more conditions? For example, if I set (1 -1) for condition contrast (task vs. baseline), is the result a effect size for r of task contrast to r of baseline? What the result represent? How to explain the results?
Thanks!
Ivan
Threaded View
| Title | Author | Date |
|---|---|---|
| Yang Yang | Apr 22, 2015 | |
| Alfonso Nieto-Castanon | Apr 22, 2015 | |
| Yang Yang | Apr 23, 2015 | |
| Alfonso Nieto-Castanon | Apr 23, 2015 | |
| Yang Yang | Apr 24, 2015 | |