Influence diagnostics are regression techniques that quantify how much individual observations affect model outputs — coefficients, predictions, and fit statistics. Parameters: each observation is individually removed or down-weighted, and the resulting change in the model is computed using measures such as DFFITS (change in fitted values), Cook's distance (aggregate influence on coefficients), DFBETAS (change in individual coefficients), and leverage (hat values, the diagonal of the hat matrix). Persistence mechanism: the procedures are codified in statistical software packages and taught as standard practice in regression analysis, persisting through documentation, software implementations, and academic curricula. [formal: influentia | substrate: behavior | horizon: hours | explicit: yes | epoch: 0.01]
Accepted ontology entry
influence diagnostics
Influence diagnostics are regression techniques that quantify how much individual observations affect model outputs — coefficients, predictions, and fit statistics. Parameters: each observation is individually removed or down-weighted, and…
Definition
Why it is in scope
Human-made set of statistical methods designed to assess whether individual observations in a dataset exert disproportionate influence on the estimated parameters or predictions of a fitted model.
Names and aliases
- influence diagnosticsen · CANONICAL
Relations from this entry
- cmsdai2d503n23vv3e00xn5baSERVES →
influence diagnostics is built for the sake of regression analysis — its designed purpose is to identify influential observations that distort regression estimates and predictions, directly serving the integrity of regression.
- cmskpjusd056nnobpc3khcmusINSTANCE_OF →
Influence diagnostics IS a specific kind of diagnostic measure: a statistical technique that quantifies how much individual observations affect regression model outputs (coefficients, predictions). It includes measures like Cook's distance, DFFIT, DFBETA. A competent statistician would classify it as a diagnostic measure. Nearest kind is diagnostic measure.
- cmskx1k6305p8nobpedwgz05sINSTANCE_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.
Relations to this entry
- cmskwle6l05nunobpbq0izi4k← INSTANCE_OF
covariance ratio is a specific diagnostic measure within influence diagnostics — it quantifies how deleting a single observation changes the estimated covariance matrix, which is one of the key measures used in influence diagnostic analysis.
Record identity
- Created
- Aug 8, 2026, 9:37 PM UTC
- Content hash
- 3d7bd0a1db8826a36f9958113667b13ce365e1a7a0a9115c012f6ec4ab9e6a8b