Because then we are unable to understand the effect of each predictor in isolation towards the target variable and our coefficients become less useful.
2/11
Checking for multicollinearity is a routine task while solving a data problem.
Most people rely on pairwise correlations for detecting multicollinearity which is a faulty approach in many scenarios
So what to do? Check this ↓
1/11
Because then we are unable to understand the effect of each predictor in isolation towards the target variable and our coefficients become less useful.
2/11
Variance Inflation refers to the inflation in the variance of the estimated coefficient of the independent variable because of the presence of multicollinearity.
3/11
A strong relationship is possible between a predictor and other variables together combined, even though there's no high correlation individually.
4/11
VIF performs a set of multivariate regression analyses to check the dependence among the independent variables by fitting multiple regression models on the dependent variables.
5/11
If we have regressors A,B,C,D, then VIF would fit models like,
A <- B,C,D
B <- A,C,D
C <- A,B,D
D <- A,B,C
and since we are looking for explained variance, we would use...
6/11
We will calculate the VIF for each predictor by using the R2 given by the model where the predictor has been used as the target.
Putting the values in the following formula we will get our VIF values:
7/11
And in the formula, high R2 would yield high VIF values as well, suggesting multicollinearity.
8/11
VIF values start from 1 and do not have an upper limit.
1 suggests that there's no multicollinearity among the variables,
as a rule-of-thumb, values > 5 or > 10 indicate high multicollinearity but not always.
9/11
