Analyses

STL decomposition

Split a series into trend, seasonal, and residual charts.

Edit on GitHub

When to use it

Use STL when you want to see the trend, the repeating seasonal pattern, and what is left, without forecasting. A forecast of the same series is ARIMA or exponential smoothing.

Assumptions

The target is numeric. Period starts at 12. The series must be at least twice that period (24 rows when period is 12). Tensr uses a robust STL fit. There is no table of component values; the output is three charts.

Running it in Tensr

Time series → Decomposition → STL Decomposition. In chat: “STL of sales, period 12.”
Target is the numeric series. Date is optional. Period starts at 12.

Options

Prop

Type

Reading the output

The same 48 months, split into trend, season, and residual with period 12. There is no forecast table.

Period = 12, N = 48. There is no component table. The three charts are the trend, the seasonal wiggle, and the residual.

Reporting (APA 7)

STL decomposition of monthly sales, N = 48, period = 12. Describe the charts; this procedure has no p-value.

Coming from SPSS

SPSS does not ship STL. Seasonal decomposition in SPSS is Analyze → Forecasting → Seasonal Decomposition (classical). Tensr’s item is Time series → Decomposition → STL Decomposition. The path string stored for this item matches that menu.

Exponential smoothing uses the same period for a forecast. Stationarity tests ask whether the series still has a unit root after you have looked at the trend.