Chapter 20. Time Series
When you have time series data, the best future prediction is usually some form of extrapolation of the historical behavior. For short-term forecasts, the behavior of interest is the short-term pattern of fluctuations. These fluctuations can be efficiently modeled by one of two methods:
regression on recently-past values
regression on past random errors
This chapter focuses mainly on these ARIMA (also known as Box-Jenkins) models, those that model a series based on lagged values of itself. Other possibilities for modeling this type of data include frequency analysis and extensions of standard regression techniques.
JMP also includes the ability to do transfer function models, where a time series is modeled using an ...