Adjusted R-Squared (Adjusted R²)

INFO

Modified R² that adjusts for the number of predictors in the model to prevent overfitting.

How It Works

$$\text{Adjusted }{R}^2=1-\left(\frac{(1-R^2)(n-1)}{n-p-1}\right)$$

• $n$: Number of observations
• $p$: Number of predictors

Adjusted R² penalizes unnecessary complexity.

What to Look For

• Use when comparing models with different numbers of predictors
• Helps detect overfitting
• Can be lower than R-Squared (R²)

Application Models

• Multiple Linear Regression
• Classification Decision Tree
• Random Forest Classification