Introduction to Linear Models

Linear models express additive effects with fixed coefficients. In strict mathematics a linear mapping also preserves scaling and has no nonzero intercept; practical straight-line regression often includes an intercept and is affine in the inputs. Such models can be useful approximations over a specified range.

A strict linear mapping preserves addition and scaling: combining inputs combines their outputs, and scaling an input scales its output. Many practical 'linear' models instead use an intercept plus constant slopes. These affine models preserve constant incremental effects but do not require zero input to produce zero output.

A useful introductory model makes its fixed and variable parts visible. It can estimate a narrow range well even when a broader process is nonlinear. Test the range and the consequences of error instead of assuming linearity because arithmetic is convenient. Known thresholds, interactions, or physical constraints may justify a different starting model immediately.

When to use it

When a simple, proportional model provides an adequate first approximation; when testing whether nonlinearity matters for a specific analysis; when budgeting and resource allocation follow approximately proportional dynamics; when understanding the baseline from which nonlinear models depart.

How it can help

Use a transparent additive or straight-line model when the task and mechanism support it. Separate fixed baselines, test the relevant range, and change the form when interactions, thresholds, or curvature materially affect the decision.