Multicollinearity & VIF: Why Correlated Predictors Break Regression

Regression textbooks warn that a Variance Inflation Factor above 10 signals trouble, but the number alone hides what's actually happening to the fit. This 2D simulator draws two predictor variables with an adjustable correlation, fits a real ordinary-least-squares model to a noisy response by inverting the normal-equations matrix from scratch, and refits it on 200 bootstrap resamples so you can watch the coefficient estimates' standard-error bars widen as the predictors become collinear — exactly when VIF explodes and R² barely moves.