JMP 18...I noticed that the "local width" option for the moving average in Smoother in Graph Builder had even numbers, which seemed strange, as moving average windows are typically odd for symmetry. So I did a few widths with Smoother (local region = fixed, local weighting = rectangular, local width = 2, 3, or 4) then compared to a Column moving average formula (equal weighting) with different n before/after.
Surprisingly, a "local width" of 4 in Smoother is equivalent to 2 before/after in Column Moving Average, so it's like (local width) / 2 + 1 which is a super weird way to define the "width" of a moving average (it's 5!).
Even weirder, how then does it handle an odd local widths? It just treats them like n - 1, except it does something strange at the ends of the time series. So there is no difference between local width 2 and 3, except at the ends.
I can't believe this would be intentional?