Validation of combine white noise using simulated data

• Main Authors: Agboluaje, Ayodele Abraham, Ismail, Suzilah, Chee, Yin Yip
• Format: Article
• Language: English
• Published: Research India Publications 2016
• Subjects: QA Mathematics
• Online Access: http://repo.uum.edu.my/21530/1/IJAER%2011%2020%202016%2010125%2010130.pdf; http://repo.uum.edu.my/21530/; http://www.ripublication.com/ijaer16/ijaerv11n20_13.pdf
• Institution: Universiti Utara Malaysia

Recent studies reveal that the data that exhibits heteroscedasticity are modelled by Exponential Generalized Autoregressive Conditional Heteroscedastic (EGARCH).Nevertheless, EGARCH model estimation is not efficient when the heteroscedasticity data have leverage effect.In this study, an algorithm is developed which is called Combine White Noise (CWN).The standardized residuals of EGARCH errors (heteroscedastic variance) are decomposed into equal variances (white noise series). The white noise series are transformed into Combine White Noise Model (CWN).The assessments of the model are based on data simulation.The simulated data of 200 and 300 sample sizes of EGARCH are generated.The generated EGARCH data are based on low, moderate and high values of leverage and skewness.Each of these generated EGARCH data is used for the estimation of EGARCH and Moving Average (MA). The same generated EGARCH data are transformed to obtain CWN data and VAR data for the estimation of CWN and VAR.Each CWN results outperformed every result of the existing models.These results confirm that CWN is the appropriate model for estimation.The CWN model fit best in the transformed 200 sample sizes of EGARCH generated data with moderate leverage and moderate skewness. While the best forecast is in the transformed 200 sample sizes of EGARCH generated data with high leverage and moderate skewness. 200 sample sizes of EGARCH generated data with right values of leverage and skewness are better than using 300 sample sizes to have reliable output.