So, I'm currently stumped by the forecasting result with a Simple Moving Average technique with period of 6-months. My time-series data has 24 data points. The calculation in question is a result from a software, and it's different with the result of my manual calculation. I have two concerns with the software's result:
- Software's result was not lagged by the MA's period. To my knowledge, if I put MA period of 6 month, the MA will start at month-7 because it needs 6 past data to calculate a single MA point. The result of this software has the MA line start directly at the first month. I need to know how the software get those early periods results.
- At the end of the data point I found out that the software's result adds its own result into the calculation. Does this sort of technique produce valid results? and is that the correct way to do it?
Actually what I really need is the validation that the software's result is valid and is following the "law" of statistics so that I could use it to do a more sophisticated forecasting techniques without doubt..
The data set in question:
[25, 22, 20, 21, 19, 23, 21, 24, 26, 24, 24, 25, 22, 23, 22, 20, 22, 22, 24, 23, 22, 22, 21, 23]
6-months period MA according to software:
[22.39, 22.91, 22.73, 22.18, 21.95, 21.36, 21.67, 21.00, 21.33, 22.33, 22.83, 23.67, 24.00, 23.33, 22.83, 23.67, 22.33, 21.83, 22.17, 22.17, 22.17, 22.50, 22.33, 22.50, 22.25, 22.13, 22.15, 22.17, 22.37]
6-months period MA according to my manual calculation:
[-, -, -, -, -, -, 21.67, 21.00, 21.33, 22.33, 22.83, 23.67, 24.00, 23.33, 22.83, 23.67, 22.33, 21.83, 22.17, 22.17, 22.17, 22.50, 22.33, 22.50, 22.25, 22.13, 22.15, 22.17, 22.37]
