Leverage

Exploring Leverage in Multivariable Linear Regression

In the context of multivariable linear regression, leverage is a distance measure that shows how far an observation is from the center of the multivariate predictor space. Observations with high leverage values would have the potential to influence the regression model highly while observations with low leverage values would not. Additionally, leverage can be used to determine if a new observation is close to the predictor space of the observations used to create the model in order to avoid extrapolation.