The last sentence in the Technical Note for tssmooth ma documentation ( [TS] pag. 586 ) reads:
It seems to me that this is not true, and the function always computes the moving-average series values, just treating the out-of-span input values as missing values, hence giving to them a coefficient of zero, according to the first sentence of the same note.
I checked this just running:
If the quoted sentence were true, sales_ma values should almost all be missing (because the first 24 values lack some lag and the last 24 values lack some lead) while they are all computed using the available values only.
Did I misunderstand something ?
Thanks
Because the first l observations and the last f observations will be outside the span of the filter,
those observations will be set to missing in the moving-average series.
those observations will be set to missing in the moving-average series.
I checked this just running:
Code:
. use http://www.stata-press.com/data/r13/sales1.dta . tssmooth ma sales_ma = sales, window(25 1 25) replace
Did I misunderstand something ?
Thanks
