Regression diagnostics are statistical methods for evaluating whether a fitted regression model meets its assumptions and identifying observations that disproportionately affect the fit. Parameters: diagnostics operate across three axes — (1) goodness-of-fit tests (e.g., residual analysis, Breusch-Pagan test for heteroscedasticity), (2) influence measures (e.g., Cook's distance, DFFITS, DFBETAS), and (3) leverage detection (hat values from the hat matrix). Each method specifies which assumption it tests or which aspect of influence it quantifies. Persistence mechanism: embedded in statistical software (R, Python statsmodels, SAS), taught in regression textbooks, and sustained through standard analytical workflows in research and industry. [formal: regressio diagnostica | substrate: behavior | horizon: hours | explicit: yes | epoch: 0.01]
Accepted ontology entry
regression diagnostics
Regression diagnostics are statistical methods for evaluating whether a fitted regression model meets its assumptions and identifying observations that disproportionately affect the fit. Parameters: diagnostics operate across three axes —…
Definition
Why it is in scope
Human-made methodology encompassing the statistical methods and visualizations used to evaluate regression model assumptions, detect influential observations, and assess model fit. It includes diagnostic measures (leverage, Cook's distance, covariance ratio) and diagnostic plots (residual plots, normal probability plots) that collectively validate regression analysis.
Names and aliases
- regression diagnosticsen · CANONICAL
Relations from this entry
- cmrxj3acr03cmsoacx73fal1oDEPENDS_ON →
Regression diagnostics needs statistics to operate — remove statistics and the methodology collapses. Object-level dependency.
Relations to this entry
- cmskflcft04ionobp9mkegvb4← INSTANCE_OF
VIF is a specific kind of regression diagnostic: it measures the inflation of coefficient variances due to multicollinearity within a regression model.
- cmskw9tqu05n4nobpl75k5tt1← INSTANCE_OF
Nearest kind: influence diagnostics is a specific kind of regression diagnostic — it measures how individual data points influence model fit. A competent speaker would call influence diagnostics 'a regression diagnostic'. Direction: influence diagnostics → regression diagnostics.
- cmsk6kqii03yxnobp2iun7hzy← DERIVED_FROM
Leverage plots (also called partial residual plots) were developed within regression diagnostics to visualize the relationship between a predictor and response while controlling for other variables. Historical: came from regression methodology, not the other way around.
- cmskb890e048unobpc31pjrms← DERIVED_FROM
Scale-location plots (spread-level plots) were developed by Cook and Weisberg (1983) within regression diagnostics to detect non-constant error variance. Historical: emerged from regression methodology.
- cmskej0ga04gdnobp1jh0r7g8← DERIVED_FROM
Added-variable plots (partial regression plots) were developed within regression diagnostics to visualize the marginal contribution of each predictor in a multiple regression. Historical: emerged from regression methodology.
- cmskrtyhp05b5nobpwflwujtw← DERIVED_FROM
Dfbetas measures the change in each regression coefficient when an observation is removed — it is derived from the methodology of regression diagnostics. Regression diagnostics existed first and fed into the computation of dfbetas.
Record identity
- Created
- Aug 8, 2026, 9:58 PM UTC
- Content hash
- c6dda1372c2aff5660d75dedf12ec8d207e9558a4c107514643ff943386bd891