Bayesian analysis of circular data in between-subjects designs
Researchers often analyze data that is either numerical, e.g. length, or is divided into (ordered) categories, e.g. level of education. However, researchers also run into data that consists of angles, measured in degrees. In that case, the data is called circular data.
This has traditionally been a field with slow development, in part due to the difficulty of deriving null distributions for test statistics. Logical solutions might include bootstrapping or, as in this project, MCMC methods in a Bayesian framework. The goal of the project is twofold: First, allow inference in a broader range of models than was previously possible for circular data. In particular, flexible GLM-type models will be developed,based on the von Mises distribution, with their associated methods for Bayesian inference, as well as models to incorporate dependence of observations in a within-between framework. Second, Bayesian hypothesis tests are developed, which provide straightforward and easily accessible methods which researchers can use to assess hypotheses that commonly occur in the analysis of circular data.
Prof. Dr. Herbert Hoijtink & Dr. Irene Klugkist
1 September 2014 – 1 September 2018