Dear Gurus,
I am new to issues of collinearity and am hoping for sage advice. From reading the forums, I know to ignore correlations among polort terms, but I'm not sure what to make of moderate correlations among regressors of interest (see below). I'm learning that orthogonalizing correlated regressors may address this issue, but how might this be implemented in AFNI? Alternatively, how does one quantify whether collinearity is an issue in their data (i.e. to answer whether moderate correlation even matters?).
Thanks in advance!
Kevin
----------- correlation warnings -----------
Warnings regarding Correlation Matrix: X.xmat.1D
severity correlation cosine regressor pair
-------- ----------- ------ ----------------------------------------
medium: -0.508 0.000 ( 0 vs. 5) Run#1Pol#0 vs. Run#2Pol#0
medium: -0.505 0.000 ( 5 vs. 10) Run#2Pol#0 vs. Run#3Pol#0
medium: -0.487 0.000 ( 0 vs. 10) Run#1Pol#0 vs. Run#3Pol#0
medium: 0.435 0.475 (19 vs. 25) cue_miss_one#0 vs. fbkMISS_one#0
medium: 0.406 0.452 (20 vs. 26) cue_miss_five#0 vs. fbkMISS_five#0
medium: 0.406 0.467 (15 vs. 21) cue_hit_zero#0 vs. fbkHIT_zero#0
medium: 0.405 0.469 (16 vs. 22) cue_hit_one#0 vs. fbkHIT_one#0