MerQuR+Statistics Glossary

Statistics Glossary → Regression coefficient (B) and parameter estimates

Regression coefficient (B) and parameter estimatesconcept

The estimated average change in the outcome produced by a one-unit increase in a predictor, holding the other predictors constant.

In the model y = b₀ + b₁x₁ + b₂x₂ + ε, b₁ is the change in the mean of y for a one-unit increase in x₁ with x₂ held constant; b₀ (the intercept) is the expected value when all predictors are zero. Each coefficient comes with a standard error (SE), a t = B/SE statistic, a p value and a confidence interval. A categorical predictor is coded with indicator (dummy) variables; each coefficient then shows the mean difference between that level and the reference level.

Because B is expressed in raw units, the Bs of predictors on different scales cannot be compared directly; standardized coefficients (β) are used for that. The sign and size of a coefficient depend on which other variables are in the model; when predictors are strongly correlated (multicollinearity), coefficients become unstable and their SEs grow. In observational data a regression coefficient is not a causal effect; it summarizes the conditional association in the model. In nonlinear models the parameters (e.g. growth rate, asymptote) do not carry this 'one-unit increase' reading and are interpreted through their role in the formula.

An everyday example

An estate agent models flat prices on floor area and building age and finds a floor-area coefficient of 28,000 TL. This means that of two flats of the same age, the one 10 m² larger is expected to cost 280,000 TL more on average. If the age coefficient is -15,000 TL, a flat of the same size is on average 15,000 TL cheaper for each additional year of age.

In MerQur

In two-way ANOVA and ANCOVA, the 'Parameter estimates' output gives a table of B, SE, t, p and CI using indicator coding with the last level as the reference. In nonlinear regression, parameter estimates with their standard errors and confidence intervals are the main output.

Analyses where this concept appears

Two-Way ANOVAANCOVANonlinear Regression

Related concepts

Standardized regression coefficient (β)Standard errorConfidence level and confidence interval