DS AdvancedLesson 24Free preview
DS Regression R-Squared
This lesson is part of Data Science · All levels
Summary
The R-squared (determination coefficient) measures the proportion of the variability of the target variable successfully described by the model, with a value range between 0 and 1. RÂ2 = 0.8 means 80% of the target variation is explained by the model and the remaining 20% by other factors. But high R2s do not guarantee good models because they can result from overfit or meaningless predictor additions.
Key points
- RÂ2 = the proportion of the target variability described by the model (0 to 1).
- RÂ2 = 0.8 means 80% of the variation is explained, the remaining 20% is another factor.
- R is always up when the predictor is added, even if it's useless.
- Adjusted R2 counts the number of predictors to be unbiased.
- Further evaluation: residual testing, adjusted RÂ2 and validation of data.
Full lesson
Code examples and quizzes
Full lessons, code examples, and quizzes are available to subscribers. 5 questions.