ABSTRACT
Time series data requires different analysis than non-time series data. Both types of data exhibit variability, but time series data can have additional structure due to the way the data is “generated” in a sequential manner. Trends, seasonality, volatility trends, and general autocorrelation in the data are all systematic components of the data that need to be considered for it to be analyzed and modeled. So, simulation of time series is accomplished by modeling the time structure and fitting an appropriate model to the randomness that remains.
This chapter explores how to decompose time series data into its systematic and random components and to conduct simulation. A variety of time series models will be examined to fit to such data, and short-term forecasts are developed. However, forecasts may also need to include patterns or variability that will differ from past patterns. We will illustrate how to include such potential change into forecasts so that lessons from both historical data, as well as deviations from such patterns, can be included in probabilistic predictions.
