Interactive time series plot: change the moving mean window, add an outlier or a structural change, or deseasonalise the data, and read trend, seasonality and irregular fluctuation from the labelled plot
A wider window is not a smoother line unless it matches the season
A time series is three things stacked on top of each other: a long-term trend, a seasonal swing that repeats on the calendar, and short irregular wobble. A moving mean is meant to peel away the swing and the wobble, but it only cancels the season when its window matches the season length. Widen the window from 6 to 9 to 12 and watch the smoothed line; then add an outlier or a structural change and watch what survives the smoothing.
Time series · season of 6
Window 1 · raw
Change the window. Watch the smoothed line between periods 6 and 42. Then change the data.
No smoothing • everything mixed
Change the data
Moving mean window1 period
136912
Window
1
Wobble removed
0%
Read more
Know This
A moving mean cancels seasonality only when its window equals the seasonal period; it smears an outlier and blurs a structural change rather than removing them.