Abstract:
|
Seasonal time series data are commonly found in environmental data, such as air pollution level, wind speed, and temperature. Notable characteristics of seasonal environmental time series data include seasonality in both mean and volatility, some irregularities in the seasonality, and possibly non-normal residuals. By considering joint mean and volatility modeling of appropriately transformed data using singular spectrum analysis, we discuss how to construct accurate point and interval forecasts with empirical quantile estimation. Further, we illustrate its application using daily maximum tropospheric ozone data.
|