We know that the correlation coefficient can take values between -1 to +1. If two variables show a very high positive correlation, then the correlation coefficient is close to one. If, at the same time, two variables show virtually no correlation, then the correlation coefficient is close to zero. These values correspond to the values of a cosine of an angle. An angle of 90 degrees has a cosine value of 0 and an angle of 0 or 360 degrees has a cosine value of 1. A 90 degrees angle sounds like two orthogonal variables and an angle of 0 or 360 degrees sounds like two overlapping variables. Can we establish a connection between a correlation coefficient and a cosine of an angle? Yes, we can, and this paper is all about that.
To read the paper go here, and to check the spreadsheet demonstrating how all the calculations are executed, click here.