Acceptable value of colinearity between regressors?

1

Dear experts,

In SPM, the orthogonality between regressors can be estimated using the design → design orthogonality tab.

Here, we found that some regressor pairs of interest had a cos theta value (colinearity) of around 0.20. Is this an acceptable value for further analysis?

Can you please point out some papers for me to use as reference?

Thanks.

2

Dear Tamer,
I’m not aware of formal decision rules, but a correlation of .2 does not seem to be a big issue. Note that problems arise when the condition number of the design matrix is large, and this number is a complex function of all correlations.
Best,
Bertrand

3

Thanks a lot for your response! It would be great if I could have a reference from the literature as well. But thank you.