Non-linearity
Non-linearity in regression arises when the true relationship between the outcome and an explanatory variable follows a curve rather than a straight line.
Also known asnonlinearity
See it move
The infographic is a two-column comparison contrasting a misspecified linear model with a correctly specified non-linear one. The left column, 'Linear model (wrong)', shows a straight fitted line that produces a U-shaped arc in the residual-versus-fitted plot — a diagnostic signal of systematic misfitting. The right column, 'Non-linear fix (right)', applies a log, polynomial, or spline form, yielding a random residual scatter and unbiased marginal effects; the caption notes that a U-shaped arc in the residual-versus-fitted plot is the standard way to detect non-linearity.