Linear discriminant analysis is a method within classification and prediction. Statistical and machine-learning methods used to assign cases to classes or estimate future outcomes, with validation against unseen data. Its usefulness depends on data structure, assumptions, validation and whether the output answers the intended decision.
Read the full entry →Research Wiki · Methods Lab
Classification and prediction
Statistical and machine-learning methods used to assign cases to classes or estimate future outcomes, with validation against unseen data.
Methods family
Statistical and machine-learning methods used to assign cases to classes or estimate future outcomes, with validation against unseen data.
Use methods only after defining the estimand, data structure, assumptions, validation plan and decision consequence.
Quadratic discriminant analysis is a method within classification and prediction. Statistical and machine-learning methods used to assign cases to classes or estimate future outcomes, with validation against unseen data. Its usefulness depends on data structure, assumptions, validation and whether the output answers the intended decision.
Read the full entry →Logistic classification is a statistical or analytical concept within classification and prediction. It should be selected for the data-generating process and decision question rather than because software makes it available.
Read the full entry →Naïve Bayes is a statistical or analytical concept within classification and prediction. It should be selected for the data-generating process and decision question rather than because software makes it available.
Read the full entry →K-nearest neighbours is a statistical or analytical concept within classification and prediction. It should be selected for the data-generating process and decision question rather than because software makes it available.
Read the full entry →Decision trees is a statistical or analytical concept within classification and prediction. It should be selected for the data-generating process and decision question rather than because software makes it available.
Read the full entry →Classification and regression trees is a method within classification and prediction. Statistical and machine-learning methods used to assign cases to classes or estimate future outcomes, with validation against unseen data. Its usefulness depends on data structure, assumptions, validation and whether the output answers the intended decision.
Read the full entry →Random forests combine many decorrelated decision trees fitted to resampled data and random predictor subsets. They often predict well, but importance measures and explanations require care.
Read the full entry →Extra trees is a statistical or analytical concept within classification and prediction. It should be selected for the data-generating process and decision question rather than because software makes it available.
Gradient boosting is a statistical or analytical concept within classification and prediction. It should be selected for the data-generating process and decision question rather than because software makes it available.
XGBoost is a regularised gradient-boosted tree algorithm that sequentially improves an ensemble by fitting residual errors. Performance depends on tuning, validation and control of leakage and overfitting.
LightGBM is a statistical or analytical concept within classification and prediction. It should be selected for the data-generating process and decision question rather than because software makes it available.
CatBoost is a statistical or analytical concept within classification and prediction. It should be selected for the data-generating process and decision question rather than because software makes it available.
Support-vector machines is a statistical or analytical concept within classification and prediction. It should be selected for the data-generating process and decision question rather than because software makes it available.
Neural networks is a statistical or analytical concept within classification and prediction. It should be selected for the data-generating process and decision question rather than because software makes it available.
Deep learning is a statistical or analytical concept within classification and prediction. It should be selected for the data-generating process and decision question rather than because software makes it available.
Ensemble learning is a statistical or analytical concept within classification and prediction. It should be selected for the data-generating process and decision question rather than because software makes it available.
Stacking is a statistical or analytical concept within classification and prediction. It should be selected for the data-generating process and decision question rather than because software makes it available.
Bagging is a statistical or analytical concept within classification and prediction. It should be selected for the data-generating process and decision question rather than because software makes it available.
Boosting is a statistical or analytical concept within classification and prediction. It should be selected for the data-generating process and decision question rather than because software makes it available.
Calibration models is a method within classification and prediction. Statistical and machine-learning methods used to assign cases to classes or estimate future outcomes, with validation against unseen data. Its usefulness depends on data structure, assumptions, validation and whether the output answers the intended decision.
Probability calibration is a statistical or analytical concept within classification and prediction. It should be selected for the data-generating process and decision question rather than because software makes it available.
ROC curves is a statistical or analytical concept within classification and prediction. It should be selected for the data-generating process and decision question rather than because software makes it available.
AUC is a statistical or analytical concept within classification and prediction. It should be selected for the data-generating process and decision question rather than because software makes it available.
Precision-recall analysis is a method within classification and prediction. Statistical and machine-learning methods used to assign cases to classes or estimate future outcomes, with validation against unseen data. Its usefulness depends on data structure, assumptions, validation and whether the output answers the intended decision.
Confusion matrices is a statistical or analytical concept within classification and prediction. It should be selected for the data-generating process and decision question rather than because software makes it available.
Sensitivity and specificity is a statistical or analytical concept within classification and prediction. It should be selected for the data-generating process and decision question rather than because software makes it available.
Lift charts is a statistical or analytical concept within classification and prediction. It should be selected for the data-generating process and decision question rather than because software makes it available.
Gains charts is a statistical or analytical concept within classification and prediction. It should be selected for the data-generating process and decision question rather than because software makes it available.
Decile analysis is a method within classification and prediction. Statistical and machine-learning methods used to assign cases to classes or estimate future outcomes, with validation against unseen data. Its usefulness depends on data structure, assumptions, validation and whether the output answers the intended decision.
Cross-validation is a statistical or analytical concept within classification and prediction. It should be selected for the data-generating process and decision question rather than because software makes it available.
Train-test splitting is a method within classification and prediction. Statistical and machine-learning methods used to assign cases to classes or estimate future outcomes, with validation against unseen data. Its usefulness depends on data structure, assumptions, validation and whether the output answers the intended decision.
Overfitting and underfitting is a statistical or analytical concept within classification and prediction. It should be selected for the data-generating process and decision question rather than because software makes it available.
Hyperparameter tuning is a statistical or analytical concept within classification and prediction. It should be selected for the data-generating process and decision question rather than because software makes it available.
Feature selection is a statistical or analytical concept within classification and prediction. It should be selected for the data-generating process and decision question rather than because software makes it available.
Feature importance is a statistical or analytical concept within classification and prediction. It should be selected for the data-generating process and decision question rather than because software makes it available.
SHAP values is a statistical or analytical concept within classification and prediction. It should be selected for the data-generating process and decision question rather than because software makes it available.
LIME explanations is a statistical or analytical concept within classification and prediction. It should be selected for the data-generating process and decision question rather than because software makes it available.
Browse linked method entries
Linear discriminant analysis
Linear discriminant analysis is a method within classification and prediction. Statistical and machine-learning methods used to assign cases to classes or estimate future outcomes, with validation against unseen data. Its usefulness depends on data structure, assumptions, validation and whether the output answers the intended decision.
Open methodQuadratic discriminant analysis
Quadratic discriminant analysis is a method within classification and prediction. Statistical and machine-learning methods used to assign cases to classes or estimate future outcomes, with validation against unseen data. Its usefulness depends on data structure, assumptions, validation and whether the output answers the intended decision.
Open methodLogistic classification
Logistic classification is a statistical or analytical concept within classification and prediction. It should be selected for the data-generating process and decision question rather than because software makes it available.
Open methodNaïve Bayes
Naïve Bayes is a statistical or analytical concept within classification and prediction. It should be selected for the data-generating process and decision question rather than because software makes it available.
Open methodK-nearest neighbours
K-nearest neighbours is a statistical or analytical concept within classification and prediction. It should be selected for the data-generating process and decision question rather than because software makes it available.
Open methodDecision trees
Decision trees is a statistical or analytical concept within classification and prediction. It should be selected for the data-generating process and decision question rather than because software makes it available.
Open methodClassification and regression trees
Classification and regression trees is a method within classification and prediction. Statistical and machine-learning methods used to assign cases to classes or estimate future outcomes, with validation against unseen data. Its usefulness depends on data structure, assumptions, validation and whether the output answers the intended decision.
Open methodRandom forests
Random forests combine many decorrelated decision trees fitted to resampled data and random predictor subsets. They often predict well, but importance measures and explanations require care.
Open methodExtra trees
Extra trees is a statistical or analytical concept within classification and prediction. It should be selected for the data-generating process and decision question rather than because software makes it available.
Open methodGradient boosting
Gradient boosting is a statistical or analytical concept within classification and prediction. It should be selected for the data-generating process and decision question rather than because software makes it available.
Open methodXGBoost
XGBoost is a regularised gradient-boosted tree algorithm that sequentially improves an ensemble by fitting residual errors. Performance depends on tuning, validation and control of leakage and overfitting.
Open methodLightGBM
LightGBM is a statistical or analytical concept within classification and prediction. It should be selected for the data-generating process and decision question rather than because software makes it available.
Open methodCatBoost
CatBoost is a statistical or analytical concept within classification and prediction. It should be selected for the data-generating process and decision question rather than because software makes it available.
Open methodSupport-vector machines
Support-vector machines is a statistical or analytical concept within classification and prediction. It should be selected for the data-generating process and decision question rather than because software makes it available.
Open methodNeural networks
Neural networks is a statistical or analytical concept within classification and prediction. It should be selected for the data-generating process and decision question rather than because software makes it available.
Open methodDeep learning
Deep learning is a statistical or analytical concept within classification and prediction. It should be selected for the data-generating process and decision question rather than because software makes it available.
Open methodEnsemble learning
Ensemble learning is a statistical or analytical concept within classification and prediction. It should be selected for the data-generating process and decision question rather than because software makes it available.
Open methodStacking
Stacking is a statistical or analytical concept within classification and prediction. It should be selected for the data-generating process and decision question rather than because software makes it available.
Open methodBagging
Bagging is a statistical or analytical concept within classification and prediction. It should be selected for the data-generating process and decision question rather than because software makes it available.
Open methodBoosting
Boosting is a statistical or analytical concept within classification and prediction. It should be selected for the data-generating process and decision question rather than because software makes it available.
Open methodCalibration models
Calibration models is a method within classification and prediction. Statistical and machine-learning methods used to assign cases to classes or estimate future outcomes, with validation against unseen data. Its usefulness depends on data structure, assumptions, validation and whether the output answers the intended decision.
Open methodProbability calibration
Probability calibration is a statistical or analytical concept within classification and prediction. It should be selected for the data-generating process and decision question rather than because software makes it available.
Open methodROC curves
ROC curves is a statistical or analytical concept within classification and prediction. It should be selected for the data-generating process and decision question rather than because software makes it available.
Open methodAUC
AUC is a statistical or analytical concept within classification and prediction. It should be selected for the data-generating process and decision question rather than because software makes it available.
Open methodPrecision-recall analysis
Precision-recall analysis is a method within classification and prediction. Statistical and machine-learning methods used to assign cases to classes or estimate future outcomes, with validation against unseen data. Its usefulness depends on data structure, assumptions, validation and whether the output answers the intended decision.
Open methodConfusion matrices
Confusion matrices is a statistical or analytical concept within classification and prediction. It should be selected for the data-generating process and decision question rather than because software makes it available.
Open methodSensitivity and specificity
Sensitivity and specificity is a statistical or analytical concept within classification and prediction. It should be selected for the data-generating process and decision question rather than because software makes it available.
Open methodLift charts
Lift charts is a statistical or analytical concept within classification and prediction. It should be selected for the data-generating process and decision question rather than because software makes it available.
Open methodGains charts
Gains charts is a statistical or analytical concept within classification and prediction. It should be selected for the data-generating process and decision question rather than because software makes it available.
Open methodDecile analysis
Decile analysis is a method within classification and prediction. Statistical and machine-learning methods used to assign cases to classes or estimate future outcomes, with validation against unseen data. Its usefulness depends on data structure, assumptions, validation and whether the output answers the intended decision.
Open methodCross-validation
Cross-validation is a statistical or analytical concept within classification and prediction. It should be selected for the data-generating process and decision question rather than because software makes it available.
Open methodTrain-test splitting
Train-test splitting is a method within classification and prediction. Statistical and machine-learning methods used to assign cases to classes or estimate future outcomes, with validation against unseen data. Its usefulness depends on data structure, assumptions, validation and whether the output answers the intended decision.
Open methodOverfitting and underfitting
Overfitting and underfitting is a statistical or analytical concept within classification and prediction. It should be selected for the data-generating process and decision question rather than because software makes it available.
Open methodHyperparameter tuning
Hyperparameter tuning is a statistical or analytical concept within classification and prediction. It should be selected for the data-generating process and decision question rather than because software makes it available.
Open methodFeature selection
Feature selection is a statistical or analytical concept within classification and prediction. It should be selected for the data-generating process and decision question rather than because software makes it available.
Open methodFeature importance
Feature importance is a statistical or analytical concept within classification and prediction. It should be selected for the data-generating process and decision question rather than because software makes it available.
Open methodSHAP values
SHAP values is a statistical or analytical concept within classification and prediction. It should be selected for the data-generating process and decision question rather than because software makes it available.
Open methodLIME explanations
LIME explanations is a statistical or analytical concept within classification and prediction. It should be selected for the data-generating process and decision question rather than because software makes it available.
Open methodNo entries match this search.