AIC & BIC: Model Comparison Criteria

AIC (Akaike) and BIC (Bayesian) Information Criteria measure model quality, penalizing for complexity to prevent overfitting.

  • Lower AIC/BIC → better model
  • AIC = n·ln(RSS/n) + 2k    BIC = n·ln(RSS/n) + k·ln(n)
  • BIC penalizes extra parameters more than AIC
  • BIC tends to select simpler models
Key rule: AIC/BIC only makes sense comparing models on the same dataset.
Try it in the app
Try: Add an X variable → re-run regression → compare new AIC to old AIC below the table. Decrease means the variable improves the model.
Watch this on our YouTube channel

Run this on your own data

Free tier, no signup needed. Works offline in the browser.

Open the app