This is a blog about psychometrics, educational testing, and a freely available R package called dexter. It is the place where we announce new versions, discuss features, and share some opinions. We try to keep the posts accessible but sometimes we do get into technical details.
dexter saw the daylight in 2017 at Cito, the Dutch National Institute for Educational Measurement, with some financial help from the Dutch ministry of education, culture, and science, and through the efforts of Gunter Maris, Timo Bechger, Jesse Koops, and Ivailo Partchev. Developed at two large players in educational testing, Cito and ACTNext, it is a serious tool for the practice of summative assessment and survey research. In particular, dexter offers:
- An easy but powerful data management system adapted to the data structure of a multi-booklet test and supporting large-scale applications
- Psychometric models consequentially developed from the scoring rules and suitable for high-stakes summative assessment
- Traditional and novel tools to monitor item quality and model fit, including a polytomous generalization of Haberman’s interaction model
- Likelihood-based and Bayesian approaches to the estimation of item and person parameters
- Novel ways to explore measurement invariance
- More than sufficient speed in handling large data
So far, there are two companion packages: dextergui (an easy graphical interface to the most important functions) and dexterMST (management, calibration and analysis of multi-stage tests).
We are thankful to Norman D. Verhelst, Robert Zwitser, and others for valuable insights, and especially to Eva de Schipper, co-author of dextergui.