I agree absolutely on analysing the residuals. In fact, this is obligatory when using linear regression in statistical inference. On the other hand, I think EDA on the explanatory variables is always useful, not for their correlations with the dependent variable (target), but to identify multi-collinearity among the explanatory variables themselves. Once identified, how you deal with it should ideally be guided by logic and domain expertise.

Love podcasts or audiobooks? Learn on the go with our new app.