What are the limitations of using R-Squared as the sole metric for evaluating the goodness of fit in a regression model?

  • R-Squared always increases with more predictors; doesn't account for bias
  • R-Squared always increases with more predictors; doesn't penalize complexity in the model
  • R-Squared is sensitive to outliers; doesn't consider the number of predictors
  • R-Squared provides absolute error values; not suitable for non-linear models
One major limitation of R-Squared is that it always increases with the addition of more predictors, regardless of whether they are relevant. This can lead to overly complex models that don't generalize well. R-Squared doesn't penalize for complexity in the model, making it possible to achieve a high R-Squared value with an overfitted model. It might not always be the best sole metric for assessing the goodness of fit.
Add your answer
Loading...

Leave a comment

Your email address will not be published. Required fields are marked *