I have seen most or all the journals I have read to use VIF and tolerance value to detect multicollinearity. But one of my colleague was suggesting to use pearson coefficient. I have run some test using same dataset and the results were very different. Which one would you use and why? or why not?