NEWS

---

mixedsubjects 1.0.0 (2026-07-02)

• Initial release.
• Seven estimators for the average treatment effect in mixed-subjects designs: Difference-in-Means, GREG, PPI++, Doubly-Tuned (D-T), Difference-in-Predictions (DiP), DiP++, and D-T DiP.
• Variance estimation via delta-method or fold-respecting bootstrap.
• Optimal design selection for allocating budget between human observations and LLM predictions.