When you have too many independent variables which are expected to be collinear too, the you have to reduce the number of the variables from the model so that the biasness from multicollinearity is removed. This tutorial provides you demonstration on one of the method which forms subgroups of variables based on their similarity in explaining the dependent variable.
This method will reduce the variables in terms of sub indices and ensure that the explaining power of the total set of variables are not reduced. This method performs well when the variables high correlation with each other and there are some theoretical sub groups.