Statistical Inference for Cultural Consensus Theory
Abstract
In the three years of support, almost all the proposed projects were completed and are published, in press, or under review in major peer reviewed journals. In particular Cultural Consensus Theory (CCT) models have been extended and/or invented for dichotomous data, ties in a graph, ordinal data, and continuous data. All the models now allow multiple consensus truths, heterogeneous item difficulty, heterogeneous informant competence, and heterogeneous informant biases. Hierarchical Bayesian inference has been provided for all of the CCT models, and software for this inference is freely available in journal articles and web sites. Bayesian post predictive tests have been developed and implemented for two critical model assumptions: (1) Are there one or more than one consensus truths?, (2) Are the items heterogeneous or homogeneous in difficulty? All of the CCT models have been applied to both simulated and real data, and it has been shown that the models perform well even with a small number of informants.
Document Details
- Document Type
- Technical Report
- Publication Date
- Feb 24, 2014
- Accession Number
- ADA605989
Entities
People
- William H. Batchelder
Organizations
- University of California, Irvine