Coefficient de détermination R²

Coefficient de détermination R²

by Thierry Ancelle

Formal & Physical Sciences Mathematics PBMathematicsPBTProbability and statistics
🎙 Thierry Ancelle 👥 25K 📅 March 3, 2017 ⏱ 13 min 👁 59K 📄 tutorial 🧭 2026-08-17
Available in: English (current) Français

Keywords

R²coefficient of determinationregressionvariancemodel fit

Summary

This educational video explains the coefficient of determination, R², in the context of linear regression. The instructor begins by cautioning against defining R² simply as the square of the correlation coefficient, as this obscures its true meaning. He then introduces a simple example with four data points to illustrate the decomposition of total variability into residual variability and variability explained by the model. The total sum of squares (SST) is calculated around the null hypothesis line (horizontal), and the residual sum of squares (SSE) around the least-squares regression line. The difference (SSR) represents the variability explained by the model. R² is defined as the proportion of total variability explained by the model, expressed as a percentage. The video covers properties of R², including its range (0 to 1) for linear regression, its relationship to the correlation coefficient, and its calculation in Excel. It also discusses interpretation, noting that a high R² indicates a strong linear relationship, while a low R² suggests other factors are at play. The concept is extended to polynomial and multivariable regression models, and the adjusted R² is introduced to account for the number of predictors. The instructor warns against using R² alone to select models, emphasizing that model choice should be based on scientific reasoning, not purely statistical fit. He illustrates this with an example where a polynomial model fits better but is biologically implausible.

229 words

Critical Evaluation

Value of the Information & Strength of the Argument

The video provides a solid conceptual foundation for understanding R², emphasizing its interpretation as the proportion of variance explained by the model. The argumentation is logical and builds from first principles, using a clear example to illustrate the decomposition of variance. The instructor effectively explains why R² is not just the square of the correlation coefficient and highlights its broader applicability. The discussion of negative R² in non-linear models is insightful and clarifies common misconceptions. The argumentation is persuasive and well-structured, making complex statistical concepts accessible without oversimplification.

Scientific Rigor, Source Quality, Title Accuracy

The scientific rigor is high; the content aligns with standard statistical theory. The instructor demonstrates expertise in statistics and epidemiology. However, the video does not cite external sources or references, relying solely on the instructor’s explanation. The title accurately reflects the content, which is focused solely on R². The video includes a brief mention of using Excel for calculations, but no formal citations. The description provides links to related videos and resources, but these are not directly cited in the video itself.

185 words

Title / Content Match

The title accurately reflects the content, which focuses exclusively on the coefficient of determination R².

Quality & Reliability

8/10

The video provides a rigorous conceptual explanation of R², grounded in the decomposition of variance, with clear mathematical definitions and practical examples. The author demonstrates expertise in statistics and epidemiology, and the content aligns with standard statistical theory. However, the video is from 2017 and does not include references to external sources, limiting its scholarly depth.

Key Moments

Markers derived by PSI from the transcript: the creator did not define chapters.

Cited Sources

Concurring Sources

Contribution & Novelties

The video provides a clear and intuitive explanation of R², emphasizing its interpretation as the proportion of variance explained by the model. It goes beyond the common definition as the square of the correlation coefficient, illustrating the concept through variance decomposition. The discussion of negative R² in non-linear models is a valuable addition, as it clarifies a common misconception. The video also highlights the importance of model selection based on scientific reasoning rather than solely on R².

Pour aller plus loin :

121 words

Radar Profile

The radar profile shows high scores in quality of information and reliability, with slightly lower scores in quantity and technical level. This indicates a focused, well-explained tutorial that may not cover all advanced aspects but provides a solid foundation.

Reliability 8/10

💬 No comments were provided for analysis.