Parametric test with REST-GCA output

Submitted by WHJUNG on
Dear REST experts, Thanks for your recent works, REST-GCA :) I am trying to perform granger causality analysis using REST-GCA. I conducted both coefficient-based and residual-based GCA for controls group and patients group and got several outputs for each group. These outputs were finally transformed into Z-value. Now, I'd like to perform between-group comparison using parametric test (i.e., two sample t-test). However, I am not sure what output files I use to do this. (1) Is it right that two outputs ("GCA_x2y_1" and "GCA_y2x_1", which are not z-transformed values) for coefficient-based GCA are used to perform between-group comparison? (2) Is it right that four outputs ("GCA_x2y_Transformed", "GCA_y2x_Transformed", "ZGCA_x2y_Transformed", and "ZGCA_y2x_Transformed") for residual-based GCA are used to compare both groups? Please comments. Thank you in advance.

YAN Chao-Gan

Sat, 12/03/2011 - 02:45

Permalink

In my opinion: 1. Coefficient-based GCA: GCA_x2y_1, .. 2. Residual-based GCA: ZGCA_*, .. You can see more information from Zhen-Xiang's recent paper. Zhen-Xiang, please give your thoughts?

The "GCA_x2y_*" and the "GCA_y2x_*" outputs of coefficient-based GCA could be used in t tests. The "ZGCA_x2y_*_transformed" and the "ZGCA_y2x_*_transformed" outputs of residual-based GCA were considered normally-distributed that could be used in t tests. It should be noted that the normality of transformed outputs of residual-based GCA is not as good as that of coefficient-based GCA. I considered that the statistic results of t tests on "ZGCA_x2y_*_transformed" and "ZGCA_y2x_*_transformed" would not be siginificant if the poplulation of your participants is small. I hope it help. Zang Zhen-xiang
Forums
Discuss