| Summary: | This study focuses on detecting a single outlier in circular data generated from a wrapped normal (WN) distribution. The discordancy tests of M, A and G 1 statistics are used to detect single outlier in simulated data generated from wrapped normal distribution. The purpose of this study is to make a comparison on the performance of these statistics via Monte Carlo simulation by obtaining the proportion of correct outlier detection for each statistic. In this study, Splus-language and R-language programming are used to carry out the simulation study. The power performance of these statistics have been investigated and the result revealed that these statistics performed better as the increment in contamination value, λ and the value of concentration parameter, ρ gets larger and close to 1 in the case of small and large sample size, n. In general, the A statistic is found to be outperformed the M and G 1statistics in all cases. As an illustration, a practical example is included in this study by using the Kuantan wind direction dataset. © Published under licence by IOP Publishing Ltd.
|
|---|