Hi,
Thank you for your help in advance. I have two questions regarding gPPI analyses in CONN.
1. I want to do correlation between the contrast (i.e., condition A > B) value from the gPPI and behavioral score outside the CONN. I see that I can do regression with any 2nd level covariate by importing the value and use calculator for each condition, but not contrast of two conditions. My questions: Should I z-score the connectivity value (gPPI contrast) before the correlation?
2. I am also using twin data and want to control the reduced population variability using permutation test. I'm not clear, but it is my understanding that the null distribution is generated with null-design within-subject (Please correct me if I'm wrong). In other tool like palm in fsl, it is possible to use "exchangeability blocks" in which the randomization is restricted in the way that twin pairs are separately randomized. However, it is second-level randomization, so the "exchangeability blocks" matters for twin data. My questions regarding this issue: (1) Is my understanding correct for within-subject randomization in CONN? (2) If so, do you think that the permutation test can control the twin data that is repeated measures?
Thank you again for your help!!
Best regards,
Hyojeong
Thank you for your help in advance. I have two questions regarding gPPI analyses in CONN.
1. I want to do correlation between the contrast (i.e., condition A > B) value from the gPPI and behavioral score outside the CONN. I see that I can do regression with any 2nd level covariate by importing the value and use calculator for each condition, but not contrast of two conditions. My questions: Should I z-score the connectivity value (gPPI contrast) before the correlation?
2. I am also using twin data and want to control the reduced population variability using permutation test. I'm not clear, but it is my understanding that the null distribution is generated with null-design within-subject (Please correct me if I'm wrong). In other tool like palm in fsl, it is possible to use "exchangeability blocks" in which the randomization is restricted in the way that twin pairs are separately randomized. However, it is second-level randomization, so the "exchangeability blocks" matters for twin data. My questions regarding this issue: (1) Is my understanding correct for within-subject randomization in CONN? (2) If so, do you think that the permutation test can control the twin data that is repeated measures?
Thank you again for your help!!
Best regards,
Hyojeong