SILPAKORN UNIVERSITY SCIENCE AND TECHNOLOGY JOURNAL, Vol 6, No 1 (2012)

Confidence Intervals for the Parameter of a Gaussian First-Order Autoregressive Model with Additive Outliers: A Simulation Study

Wararit Panichkitkosolkul, Luckhana Saothayanun, Yupin Kanjanasakda, Sunee Taweesakulvatchara

Department of Mathematics and Statistics, Faculty of Science and Technology, Thammasat University, Phathumthani

Department of Applied Statistics, Faculty of Science, University of the Thai Chamber of Commerce, Dindaeng, Bangkok

Department of Applied Statistics, Faculty of Science, University of the Thai Chamber of Commerce, Dindaeng, Bangkok

Department of Applied Statistics, Faculty of Science, University of the Thai Chamber of Commerce, Dindaeng, Bangkok


Abstract


This paper is concerned with interval estimation of a parameter for a Gaussian first-order autoregressive model, AR(1), when there are additive outliers in a time series. We compared the confidence intervals based on the weighted symmetric estimator (), the recursive mean adjusted weighted symmetric estimator (), the recursive median adjusted weighted symmetric estimator (), and the improved recursive median adjusted weighted symmetric estimator () by using Monte Carlo simulation. Simulation results have shown that the confidence interval based on the estimator is better than the other confidence intervals with respect to the coverage probability and the length criteria.

Key Words: AR(1) model; Additive outliers; Confidence interval; Coverage probability; Length

Full Text: PDF