New Zealand visitors, 1980 to 2000
Use this example to show how an additive model can fail to fit a data set.
For the additive model, click Trend and then Seasonal to add a quadratic trend and seasonal effect. Observe that the actual seasonal pattern changes in amplitude, so forecasts would under-estimate tourist numbers in the NZ summer and over-estimate them in the winter.
Select Multiplicative model. This initially displays the logarithms of the numbers. The amplitude of the seasonal effect seems more constant.
Click Trend and then Seasonal to find the trend and seasonal effect on the log scale (i.e. fit a multiplicative model). Then use the slider to remove the log transformation. This model implies:
This example describes the numbers of short-term visitor arrivals in New Zealand (in thousands) each quarter from 1980 to 2000.