What should be the correlation threshold value based on which we determine the highly collinear variables?
Answers
solution:
If two predictor variables X1 and X2 have a correlation of above about .70, this suggests that one of two things may be occurring.
• X1 and X2 may be two different ways of measuring the same thing.
• X1 and X2 are so strongly confounded that their predictive contributions will be very difficult to separate.
If X1 and X2 are just two measures of the same ‘construct’, the decision what to do is relatively easy. You may be able to combine or average these variables into one measure, or select just one of them to represent the construct.
If X1 and X2 clearly represent different constructs, but they are strongly confounded (highly correlated), their effects are very difficult to separate no matter what analysis you use. The decision what to do in that case is more difficult.
If X1 and X2 are highly correlated with each other, they provide redundant information; they will compete to explain a lot of the same variance in Y.