SECTION 13.3 Summary

• Time series often display a long-run trend. Some time series also display a strong, repeating seasonal pattern.
• Regression methods can be used to model the trend and seasonal variation in a time series. Indicator variables can be used to model the seasons in a time series. When indicator variables are used in a regression applied to untransformed data, the indicator variables capture additive seasonal effects.
• Transformations, such as the logarithm, can simplify the regression modeling process for trend and seasonal fitting. Seasonal indicator variables used in the modeling of logged data capture multiplicative seasonal effects.
• Examine the residuals from a regression-based time series model to see if there is any evidence of systematic patterns, such as autocorrelation, that are not adequately captured by the model.