Régression logistique.  2) Pratique

Régression logistique. 2) Pratique

🎙 Thierry Ancelle 👥 25K 📅 December 21, 2022 ⏱ 29 min 👁 7K 📄 tutorial 🧭 2026-08-17
Available in: English (current) Français

Keywords

logistic regressionodds ratiovariable selectionlikelihoodAICBICinteractioncolinearity

Summary

This video is the second part of a course on logistic regression, focusing on practical aspects. The instructor, Thierry Ancelle, explains how to select variables for multivariable analysis, interpret output tables, and perform stepwise model reduction using likelihood ratio tests. He illustrates the process with a case study on toxoplasmosis seroconversion in pregnant women. The video covers variable coding for quantitative, ordinal, and categorical variables, including the use of indicator variables. It also discusses interaction terms and how to handle them. The instructor demonstrates commands in R and Stata, and highlights common problems such as automatic procedures, poor coding, small sample sizes, and collinearity. The video concludes with a discussion of information criteria (AIC and BIC) for model selection, emphasizing the importance of checking significance of likelihood differences. Overall, it provides a comprehensive practical guide for researchers in epidemiology and biostatistics.

141 words

Critical Evaluation

Value of the Information & Strength of the Argument

The video provides valuable practical guidance on logistic regression, with a clear step-by-step approach. The argumentation is solid, as the instructor explains the rationale behind each step, such as why certain variables are included or excluded, and how to interpret results. The use of a real-world example (toxoplasmosis study) enhances the practical value. The explanation of likelihood and deviance is clear, and the demonstration of model reduction using likelihood ratio tests is instructive. The discussion of information criteria and their limitations is balanced. The video also addresses common pitfalls, such as collinearity and interaction, which are often overlooked in introductory courses.

Scientific Rigor, Source Quality, Title Accuracy

The scientific rigor is high, as the content is based on established statistical methods. The instructor does not cite specific external sources, but the methods are standard in epidemiology. The title accurately reflects the content. The video is well-structured and the explanations are precise. However, the lack of references to literature or further reading is a minor weakness. The video is suitable for an audience with some background in statistics, but it is not overly technical.

192 words

Title / Content Match

The title accurately reflects the content, which focuses on the practical application of logistic regression.

Quality & Reliability

8/10

The video is a well-structured tutorial on logistic regression, presented by an expert in epidemiology. It covers practical aspects with clear examples and explanations. The content is scientifically sound, but it does not provide references to external sources, and the presentation is primarily pedagogical rather than research-oriented.

Key Moments

Cited Sources

Concurring Sources

Contribution & Novelties

The video offers a practical, hands-on approach to logistic regression, emphasizing variable selection, model reduction, and interpretation. It provides a clear example with a real dataset, which is valuable for learners. The explanation of likelihood and deviance is accessible, and the step-by-step demonstration of model building is instructive. The video also covers important topics like interaction and collinearity, which are often not covered in introductory courses.

Pour aller plus loin :

119 words

Radar Profile

The radar profile shows high scores in quantity and quality of information, with a moderate technical level. The reliability is high, indicating a trustworthy educational resource. The video is well-balanced, with strengths in providing detailed practical guidance and clear explanations.

Reliability 8/10