I had some trouble with the ETS tool on this because i've never used that before (and i'm using 11.5, so the output didn't seem very intuitive to me). I got to that point at around the 9 minute mark, so i'm pretty happy with myself, and I think the results are right in line with expectations:
[Need help] Could someone tell me about nonlinearity in Time Series?
Alexeryx version 2019.2.5.62427 - installed in the morning
ETS (Summer1 + Summer2) = 702.87
ETS (Summer1) + ETS (Summer2) = 356.67 + 339.68 = 696.35 <> ETS (Summer1 + Summer2)
On my mind the forecast of a sum it should be equal the sum of a forecasts.
For linear regression it is always true - linear regression could be shown as a kernel function.
I suggest that may be in two cases:
- because ETS returns not the Expected Value but 50% quantile - in this case I am affraid if I could relly in Alteryx forecasting module in my work.
- in TS Alteryx take different periods for Summer 1 and Summer 2 like Summer1 = sin (year / 5) and Summer2 = sin (year/7) and it find no single valid harmonic in the sum. It is guess - I know nothing about TS.
Here is my solution. A little more lengthy than others. It is my first Alteryx workflow and my brain is still approaching these problems from a stored procedure point of view (which is a habit I recognize will dissipate over time...). Cheers!