Abstract:
In this paper, a new definition of circular residuals based on circular distance is proposed. The property of the new residuals enables us to perform two tasks: the diagnostic checking on the assumption of linear circular regression model and the identification of outlier in a circular regression. The identification is made by using several graphical and numerical methods including PP plot, QQ plot, circular plot, C-statistic, D-statistic and M-statistic. Numerical and simulation studies show that the new residuals performed well for the above task