multicollinearity
Multicollinearity occurs in a regression model when two or more predictor variables are strongly linearly related, so they contain overlapping information. The model may still make useful predictions, but it can become difficult to determine each predictor's separate contribution: estimated coefficients may be unstable, change substantially with small changes in the data, or have larger standard errors. This is especially important when the goal is to interpret coefficients, not just predict outcomes. Analysts may investigate correlations and variance inflation factors, then consider combining or removing redundant predictors, collecting more informative data, or using regularization. The right response depends on the task; a high correlation among predictors is a diagnostic concern, not automatically a reason to discard a variable.