What's so bad about collinearity?
To start with, it is the matter of collinearity not correlationThe two are connected, however not indistinguishable. Collinearity includes sets of variables, not really only two (as in correlation). For instance, on the off chance that you had 10 free variables that were totally uncorrelated and one that was the entirety of those 10, you would have no high correlations however you would have collinearity.
Next, this presents a few issues.
Outrageous affectability to change in input information. In certain circumstances you can roll out a modest improvement in the information and end up with totally unique relapse conditions
High fluctuation of the parameter gauges - that is, the parameter evaluations won't be all around assessed. This can prompt no variables being viewed as critical.
Wrong signs on the parameters.
It can bring issues into robotized variable determination strategies. All the stepwise/in reverse/forward techniques are immensely effective even without collinearity, however better strategies, for example, LASSO can run into issues when it occurs