On July 13, ALSET statistician Teck Kiang Ten presented a paper titled “Doubly Classified with R” at useR! 2018, an annual conference for aficionados of the programming language R. The paper follows Tan’s publication of a new book on doubly classified models in November 2017.
Tan’s presentation began with an overview of doubly classified models, which are used to analyze contingency tables that are doubly classified. These models aim to reveal patterns out from doubly classified tables using log-linear models.
“Analyzing changes over two time points of a common variable is common for panel data analysis and doubly classified models are well suited for carrying out such analysis in many different fields,” says Tan. “Social science researchers might use them to examine the patterns of intergeneration social mobility, inter-rater reliability, or intermarriage between ethnic groups. Clinicians might use them to analyze associations between the right and left eyes.”
Tan pointed out that doubly classified models are not new, but standard textbooks cover only a few of them, and journal articles usually describe them in technical language that is difficult for those with little mathematical and statistical background to understand. His talk focused on conveying these models using a new graphical approach called symbolic tables that makes them more accessible. Using a few standard R functions, he illustrated the procedure of setting up a doubly classified model for the complete symmetry model and the odds symmetry model I and II.
Tan also introduced his recently written book on this topic, Doubly Classified Model with R. The book covers doubly classified modeling strategies, the symbolic tables for the 44 doubly classified models, and the principle guide to build one’sown model. The R codes are storedin his personal web page at the Microsoft Azure JNB.
The was organized by Monash University and held at the Brisbane Convention and Exhibition Centre in Brisbane, Australia. The presentation can be found on YouTube via this link.