Can we detect multicollinearity with correlation (cooeficient) ? Or it has to be based on conflicting f-stat, tstat and r2?
High Fstats, and low t values (the coefficients are not significant) and usually there is HIGH Correlation between the 2 independent variables.
yeah you can use the correlation rule (greater than 70% means there is multicollinearity) if there are ONLY two independent variables. Otherwise, look at the Fstats and Tvalues.
a rule of thumb, if absolute correlation is higher than 0.7, Multicollinearity is present. That’s what a quant analyst said.