The widely applicable information criterion

Predictive evaluation asks how well a model predicts observations that were not used for fitting. We compare models using loss functions and cross-validation.

Available Wednesday, October 14, 2026.

---
title: "The widely applicable information criterion"
toc: false
course-release-id: model-criticism-and-prediction
course-release-placeholder: true
---

Predictive evaluation asks how well a model predicts observations that were not used for fitting. We compare models using loss functions and cross-validation.

**Available Wednesday, October 14, 2026.**