Multicollinearity occurs when two or more predictor variables are highly correlated. In regression analysis, their estimated coefficients can become unstable and change with minor data differences, making individual effects difficult to interpret. section 3.4.3
Adding more variables does not automatically improve a model: too many predictors can muddy the analysis, contribute to multicollinearity, or overfit noise instead of revealing underlying patterns. section 3.4.3
Review-score regression
A review-score model might include delivery time, price, freight value, payment value, product attributes, and delivery-time misestimation. If predictors overlap strongly, treat their coefficients cautiously rather than claiming that one variable alone explains customer satisfaction. section 3.4.3
