Assessing a Bayesian Embedding Approach to Circular Regression Models

Abstract

Circular data is different from linear data and its analysis therefore also requires methods that are different from the conventional methods. In this paper the Bayesian embedding approach to estimating circular regression models as presented in Nu˜nez-Antonio, Guti´errez-Pe˜na, and Escarela (2011) will be investigated, by means of five simulation studies, in terms of performance, efficiency and flexibility. In addition an empirical example of a regression model predicting teachers’ scores on the interpersonal circumplex will be used throughout the paper. From the simulation studies we may conclude that performance is reasonable in most situations and the method is deemed efficient and very flexible. Researchers should however take care when using this method on extreme data and with the interpretation of their results.