R2 Goodness of Fit

Also called the coefficient of determination.

$$R^2=1-\frac{{\sum }_{i=1}^N(y_i-{\hat{y}}_i{)}^2}{{\sum }_{i=1}^N(y_i-\bar{y}{)}^2}$$

Note that $0\le R^2\le 1$, $R^2=1$ : Perfect regression. $R^2=0$ : Baseline model. Predictions are the averge.

You can think of it as the ”% of dependent variable variations that the linear model explains”.

Note that $R^2$ does not indicate whether the linear regression model is “good enough”. To see this, you must evaluate the statistical significance of each coefficient.