Linear regression is inappropriate to model binary responses such as pass/fail, alive/dead. Learn the principle of logistic regression, a generalized linear model, its similarities with linear regression and its specific tools. Common pitfalls and best practices are discussed.
In this course, the construction of regression models for categorical data is covered with the help of real case studies.
Course Outline
Session 1: Modelling a Binary Response
- Logistic Regression (Generalised Linear Models)
- Odds ratios
- Significance testing
- Assessing the performance of the model
- Using the model for prediction purposes
Session 2: Modelling a Response with More than 2 Levels
- Ordinal Regression
- Multinomial or Polytomous Regression
- Significance testing
- Assessing the performance of the model
- Using the model for prediction purposes
Course Duration
The recommended course duration is 2 online sessions.
Target Audience
This module is intended for anyone who collects binary, categorical and/or ordinal data and makes decisions based on this data. This regression technique will be particularly useful for people who have to explain qualitative data and work in finance, epidemiology, medicine, genetics, human sciences, econometrics, marketing.
Participants must have completed the following modules or have an equivalent level of knowledge: