Time Series Analysis Exponential Smoothing Methods 2 — Questions and Answers
Question 1: What is 'damped trend' exponential smoothing?
- A method that dampens the trend toward a flat forecast over longer horizons using parameter φ (Correct answer)
- A method that removes trend entirely
- A method where the trend increases over time
- Simple exponential smoothing with no trend
Correct answer: A method that dampens the trend toward a flat forecast over longer horizons using parameter φ
Damped trend adds a damping parameter φ (0 < φ < 1) that causes long-horizon forecasts to flatten out rather than extrapolate the trend indefinitely.
Question 2: How are the smoothing parameters (α, β, γ) typically estimated in exponential smoothing models?
- By minimizing the sum of squared one-step-ahead forecast errors (Correct answer)
- By setting them manually to fixed values like 0.5
- By solving a system of linear equations
- By using cross-validation on test data only
Correct answer: By minimizing the sum of squared one-step-ahead forecast errors
Parameters are optimized numerically to minimize in-sample forecast error, most commonly the sum of squared errors (SSE).
Question 3: Which Python class in statsmodels implements the full Holt-Winters exponential smoothing model?
- ExponentialSmoothing (Correct answer)
- HoltWinters
- ets()
- SimpleExpSmoothing
Correct answer: ExponentialSmoothing
statsmodels.tsa.holtwinters.ExponentialSmoothing provides the full ETS model including additive and multiplicative trend and seasonality options.
Question 4: What is the 'initialization' problem in exponential smoothing?
- The choice of starting values for level, trend, and seasonal components affects early forecasts (Correct answer)
- The model cannot start without exactly 2 years of data
- Parameters cannot be estimated before smoothing begins
- The first forecast is always zero
Correct answer: The choice of starting values for level, trend, and seasonal components affects early forecasts
Exponential smoothing requires initial values for the state components; poor initialization can lead to inaccurate early predictions and suboptimal parameter estimates.
Question 5: Why can multiplicative error ETS models produce prediction intervals that are asymmetric?
- Because multiplicative errors scale with the level, creating non-constant forecast variance (Correct answer)
- Because the level parameter is always greater than 1
- Because seasonal indices are always positive
- Because the trend is multiplicative
Correct answer: Because multiplicative errors scale with the level, creating non-constant forecast variance
When errors are multiplicative, the forecast variance grows with the forecast level, producing wider intervals at higher levels and narrower ones at lower levels.
Question 6: For which type of series is simple exponential smoothing most appropriate?
- Non-seasonal data with no systematic trend (Correct answer)
- Data with strong seasonality
- Data with a clear linear trend
- Data with multiple seasonal cycles
Correct answer: Non-seasonal data with no systematic trend
SES is designed for series that fluctuate around a gradually changing level without a consistent trend or seasonal pattern.
Time Series Analysis Exponential Smoothing Methods 2
What is 'damped trend' exponential smoothing?