Course objective: Introducing the logistic linear regression model for two or more class probabilities. The scope of methods covers the maximum-likelihood estimation of the model, the tests for significance of effects, the search for the best subset of explanatory variables and the prediction of a class label using the model.

Organization: The course is divided into five sessions. Each session is made of a lecture and practical classes. In those in-class practical classes, students are given the opportunity to tackle problem-solving exercises using R.

The following documents are provided for each session:

Past exams:

To start with Rmarkdown: brief introduction.

Session 1

Objective: A post on the blog of a French radio channel reports that the risk of being infected by COVID19 is 33% lower in people with blood type O. What does this assertion means?

The session introduces a statistical method to compare groups of items regarding their exposition to a given risk.

Session 2

Objective: How to use colorimetry to evaluate the maturity of a fruit? The session introduces a mathematical framework and related statistical methods to address this issue.

Session 3

Objective: Are the components of a colorimetry profile all equally important to predict the maturity of a fruit? The session introduces the test for the significance of effects in the general setup for modeling class probabilities introduced in Session 2.

Session 4

Objective: How to choose the best subset of explanatory variables to explain class probabilities in a large profile of possibly correlated features?

Session 5

Objective: Prediction of the class label of test items exposes to errors. Those error rates need to be accurately evaluated, in order to help designing optimal classification strategies.