Exponential smoothing
Forecasts a series by exponentially weighted averaging — simple (level only) or Holt (level and trend).
Menu: Analyze → Forecasting → Exponential Smoothing…
When should I use it?
- Short-term forecasts of sales or demand without a strong seasonal pattern.
What are the assumptions?
- Equally spaced observations; no seasonality (decompose first if needed).
How do I run it in StatLab?
- Choose the series and the model; leave alpha/gamma blank to estimate them.
- Set forecast periods and optionally save fitted values.
- For seasonal data choose Winters' additive (constant seasonal swing) or multiplicative (swing grows with the level) and give the periodicity.
How do I read the output?
| Output | Meaning |
|---|---|
| Model Parameters | Alpha (level) and gamma (trend) weights, SSE and RMSE. |
| Fitted Values | One-step-ahead forecasts and errors. |
| Delta (season) | Weight given to the newest seasonal information (Winters models). |
Worked example
On the practice survey dataset (File → Sample dataset), this command — which the dialog's Paste button also produces — runs the procedure:
PROC tsSmoothing /v=score /model=simple /period=12 /forecast=0 /save=NO.
Output tables: Model Parameters · Fitted Values.
Computed by StatLab's validated procedure code when this page was built. The example only illustrates the steps; choose variables that fit your own research question.
What are common mistakes?
- Using the simple model on a trending series — it lags behind.