What does multicollinearity in regression refer to?

Multicollinearity in regression specifically refers to high correlations among input variables. When two or more independent variables in a regression model are highly correlated, it can create issues in estimating the relationships between the variables and the dependent variable. This phenomenon can lead to inflated standard errors for the coefficients, making it difficult to determine the individual impact of each predictor. As a result, multicollinearity can affect the statistical significance of the predictors and may complicate the interpretation of the model.

Understanding multicollinearity is crucial for regression analysis because it can also lead to unwanted behavior in model predictions and overall performance. Identifying and addressing multicollinearity, such as by removing or combining correlated variables, is important in building a robust regression model.

In contrast, the other choices refer to different concepts. High variance in the target variable pertains to the variability of the dependent variable rather than the relationships among independent variables. Non-constant variance of the target variable relates to heteroscedasticity, which is an issue in regression models that affects the validity of inference, and high skewness in input distributions refers to the asymmetry of the variable distributions, which is another aspect of data quality but is not directly related to multicollinearity.