Słownik AI
Kompletny słownik sztucznej inteligencji
Sensitivity Analysis
Systematic method evaluating how variations in input parameters affect model results to identify the most influential factors.
Robustness Test
Statistical procedure verifying the stability of analysis conclusions when underlying assumptions or data are modified.
Leverage Points
Observations with extreme predictor variable values that exert disproportionate influence on regression model fitting.
Cook's Distance
Quantitative measure combining leverage and residuals to identify observations having excessive impact on estimated regression coefficients.
Influence Analysis
Set of statistical diagnostics quantitatively identifying individual observations significantly affecting model parameters and predictions.
Collinearity Diagnostic
Procedure detecting high correlations between predictor variables that may compromise the stability and interpretability of regression coefficients.
Structural Break Test
Statistical analysis checking whether model parameters remain constant across different periods or data subsets.
VIF (Variance Inflation Factor)
Indicator quantifying the increase in coefficient variance due to multicollinearity, with values above 10 indicating serious problems.
Jackknife method
Systematic resampling technique that sequentially removes each observation to assess the sensitivity of estimators to individual observations.
Normality test of errors
Statistical verification of the distribution of residuals to ensure that the underlying inference assumptions of the model are met.
Ramsey RESET test
Specification diagnostic testing whether the functional form of the model is appropriate by detecting nonlinearities or omitted variables.
Coefficient stability
Evaluation of the constancy of estimated parameters when data or model specifications are changed, measuring the robustness of inferences.
Winsorization methods
Technique for treating extreme values by replacing observations beyond certain quantiles with the values of those quantiles to reduce their influence.