Discriminant Analysis Click on a title to access the corresponding entry of the Glossary Click on a number to access the detailed Table of Contents of the Tutorial

 CLASSIFICATION   Decision Trees Discriminant Analysis Fisher's discriminant Logistic Regression

 Tutorials 1 Overview of Decision Trees   What are Decision Trees ?    Growing the Tree.    Choosing a predictor for a split.    How big a Tree ?    Using the Tree. 2 Different types of predictors   Categorical predictors :        * Merging the modelities ?        * Special case : binary dependent variable.    Ordinal predictors : merging and splitting.    Numerical predictors.
 Tutorials 3 Splitting nodes   Impurity criterion :       * Misclassification rate.       * Gini index.       * Entropy.       * Variance.    Chi-square based splitting : CHAID    The "Twoing" split.    Priors, weights and costs. 4 The "right size" Tree     Tree overgrowing.     Stopping the growth.     Pruning a large Tree.

 Tutorials 1 Overview of Discriminant Analysis    Discriminant Factor Analysis.     Building a classifier.     Assumptions on classes.     Choosing the best set of predictors. 2 Discriminant Factor Analysis    Which variables discriminate between classes ?    Creating discriminant dunctions.    Is the model trustworthy ?    Predictor selection.    Categorical variables.
 Tutorials 3 The classifier    Geometric discriminant function.    Probabilistic discriminant functions.    Building the classifier.    Is the model trustworthy ?    Predictor selection.    Categorical variables. Complements    Mahalanobis distance. Metrics.    Mahalanobis distance criterion.   Decomposition of the total covariance matrix.    Variance of a projection on a line.

 Tutorial 1 Fisher's criterion J.    Fisher's vector.    Is Fisher's discriminant optimal ? 2 Mathematical complements :        * Derivative of a quadratic form.        * Finding the maximum of the ratio of two quadratic forms.
 Interactive Animation Classes can be dragged around.     Line can be spinned to Fisher's position.

 Tutorial 1 Intuitive approach to Logistic Regression :        * Regression on class indicators.        * The logistic function.    From Discriminant Analysis to Logistic Regression :        * The logit function.        * Direct calculation of posterior probabilities.
 Tutorial 2 Estimating the parameters :       * Likelihood maximization.       * Non-linear optimization.       * Local minima.       * Special case : classes are separable.

___________________________________________________________________________________