Time Series Data

A time series is an ordered collection of measurements taken at regular intervals--for example, daily stock prices or weekly sales data. The measurements may be of anything that interests you, and each series can generally be classified as one of the following:

• Dependent. A series that you want to forecast.
• Predictor. A series that may help to explain the target--for example, using an advertising budget to predict sales. Predictors can only be used with ARIMA models.
• Event. A special predictor series used to account for predictable recurring incidents—for example, sales promotions.
• Intervention. A special predictor series used to account for one-time past incidents--for example, a power outage or employee strike.

The intervals can represent any unit of time, but the interval must be the same for all measurements. Moreover, any interval for which there is no measurement must be set to the missing value. Thus, the number of intervals with measurements (including those with missing values) defines the length of time of the historical span of the data.