Recomposing Data with iNZight
Recompose the time series data.
Trend line + Estimated Seasonal Effect
If the seasonal component increases in range over the time series, a multiplicative model would be better.
iNZight is an additive model, so it suggests that the early seasonal cycles are below average, and later seasonal cycles are larger than average.
This can be seen by the Residuals being larger at either ends of the time range, and less in the centre.
(Residuals are more negative at the start and more positive at the end)
Another example of this:
Total numbers of visitors to NZ:
The seasonal component increases in range from 1999 to 2002, then similar seasonal cycles from 2003 to 2009.
There is a slight increase in seasonal cycles 2010-12 with a much higher visitor numbers around September 2011
Analyse the Recomposed data for these graphs: (Tourist numbers to NZ from different countries of origin)