Residual

A residual is the difference between an observed value and its model-predicted value. Residuals reveal where and by how much a model misses its data.

Observed value: A value recorded through measurement or observation. It supplies the data point from which a residual is calculated.
Statistical model: A mathematical representation of patterns in data, often used for explanation or prediction. A residual is defined relative to a model's prediction.
Model diagnostics: Methods for checking whether a fitted model adequately represents data and satisfies its assumptions. Residual patterns are among the principal evidence used in diagnostic checks.
Prediction error: The difference between a predicted value and the value that actually occurs. It is a broad term; residual commonly names this difference for observed fitting data.
Predicted value: A value estimated by a model for a specified case or input. Subtracting this estimate from the observation gives the residual.
Regression analysis: A method for modeling the relationship between a response variable and one or more predictors. Regression models provide a standard setting for calculating residuals.
Heteroscedasticity: A condition in which a model's error variance changes across observations or predictor values. A funnel-shaped residual pattern can indicate nonconstant variance.
Deviation: The difference between a value and a chosen reference, such as a mean or prediction. A residual is a specific deviation measured from a model prediction.
Fitted value: A model's prediction for a data point used to estimate its parameters. Residuals compare each fitted value with its corresponding observation.
Least squares: A parameter-fitting method that minimizes the sum of squared residuals. Its fitting criterion gives residuals a central role in estimating model parameters.

Linked from 17 pages