nearcollinearity

Nearcollinearity is a term used in statistics and data analysis to describe a situation where predictor variables in a regression model are almost linearly dependent. This means that one or more predictor variables can be almost perfectly predicted by a linear combination of the other predictor variables. While not a true case of perfect collinearity (where variables are exactly linearly dependent), nearcollinearity can still cause significant problems in regression analysis.

The primary issue with nearcollinearity is that it inflates the standard errors of the regression coefficients. This makes it difficult to determine the individual statistical significance of the affected predictors. Coefficients might appear insignificant even when the predictor itself is theoretically important. Furthermore, small changes in the data or the model specification can lead to large and unstable changes in the estimated coefficients. This lack of stability makes it hard to interpret the model and draw reliable conclusions about the relationships between the predictors and the response variable.

Detecting nearcollinearity often involves examining variance inflation factors (VIFs). A high VIF for a particular predictor indicates that it is highly correlated with other predictors in the model. While there is no universal threshold for what constitutes a "high" VIF, values above 5 or 10 are often considered indicative of problematic nearcollinearity. Addressing nearcollinearity might involve removing redundant predictors, combining variables, or using regularization techniques like Ridge regression.