I attended the 2019 Northeastern Educational Research Conference at Trumbull, CT. The conference is held at Trumbull Marriott Shelton annually. My session is titled Estimation: New Directions and Method Comparisons.
My research project is about "A Hierarchical Mixed Membership Model for Multiple-Answer Multiple-Choice Items with Signal Detection Theory."
Other presenters in my session include Dakota Cintron (On the Comparative Performance of Iterative, Variational and Stochastic Estimation Methods in Latent Class Analysis), Zachary Collier (Cross-Validation and Bootstrap to Evaluate the Number of Subgroups in Finite Mixture Model), and Paulius Satkus (An Applied Example of Two-Tier Multiple-Group Testlet Model). We discussed many model comparison problems in terms of estimation algorithm and testing.
I received many useful feedbacks from the audiences, including:
- how to extend the application of signal detection theory for other psychometrician;
- how to compare and link the signal detection theory with other conventional measurement models (e.g., cognitive diagnosis models and item response theory);
- how to improve the speed and accuracy of Bayesian estimation in latent class models.
I also attended a special seminar by Alina von Davier (senior vice president of ACTNext), Rochelle Michel (NERA president), and Ellen Forte (CEO and Chief Scientist of edCount). They shared their stories as educational practitioners and researchers.
Here are the presentation slides: