A Monte Carlo Simulation Comparison of Methods of Detecting Outliers in Time Series Data

Authors

Egbo. Mary Nkechi, Department of Statistics, Chukwuemeka Odumegwu Ojukwu University, Uli, Anambra State, NigeriaFollow
Bartholomew. Desmond Chekwube, Department of Statistics, Federal University of Technology Owerri, Imo State, NigeriaFollow
Obite. Chukwudi Paul, Department of Statistics, Federal University of Technology Owerri, Imo State, NigeriaFollow
Kiwu. Lawrence Chizoba, Department of Statistics, Federal University of Technology Owerri, Imo State, NigeriaFollow

Abstract

Observations that are considerably different from the rest of the data are referred to as outliers. Outliers in a dataset have a number of undesirable consequences for statistical analysis. The negative implications could include a decline in prediction quality and the inclusion of mistakes in model parameter estimates. Currently, just a few literature reviews have been done on these topics. As a result, four outlier detection methods that are specifically developed to find outliers in univariate standard normal time series datasets were compared. Comparative approaches that are simple like Mean Squared Error, Coefficient of Variation, Standard Error of Mean and Percentage Mean Success Rate, computed from outliers detected in a monte carlo simulation of samples of sizes 500 and 1,500 were proposed and used to select the best outlier detection method(s).

Digital Object Identifier (DOI)

http://dx.doi.org/10.18576/jsap/110306

Recommended Citation

Mary Nkechi, Egbo.; Desmond Chekwube, Bartholomew.; Chukwudi Paul, Obite.; and Lawrence Chizoba, Kiwu. (2022) "A Monte Carlo Simulation Comparison of Methods of Detecting Outliers in Time Series Data," Journal of Statistics Applications & Probability: Vol. 11: Iss. 3, Article 6.
DOI: http://dx.doi.org/10.18576/jsap/110306
Available at: https://digitalcommons.aaru.edu.jo/jsap/vol11/iss3/6