Forecasting Flashcards
7 cards from real CBE practice questions. Tap to flip, then mark Knew It or Still Learning — missed cards come back until you master them.
Read the first 7 Forecasting flashcards as text
The Box-Jenkins methodology for ARIMA modeling follows which sequence of steps?
Answer: Identify → Estimate → Diagnose → Forecast
Box-Jenkins proceeds by identifying the model order via ACF/PACF, estimating parameters, diagnosing residuals, and then forecasting if diagnostics pass.
The Partial Autocorrelation Function (PACF) is primarily used in ARIMA modeling to identify:
Answer: The order of autoregressive (AR) terms
The PACF cuts off sharply after lag p for an AR(p) process, making it the standard diagnostic for selecting the autoregressive order.
A forecast combining multiple models typically outperforms individual models because:
Answer: It leverages diverse information and reduces variance without proportionally increasing bias
Forecast combination diversifies across different model structures and information sets, reducing overall error variance through a bias-variance trade-off.
Which statement about out-of-sample forecast evaluation is correct?
Answer: Out-of-sample evaluation uses data not seen during model estimation
Out-of-sample testing uses a reserved holdout period after the estimation window to provide unbiased evidence of how the model will perform on new data.
When MAPE (Mean Absolute Percentage Error) is undefined or problematic, the most likely reason is:
Answer: Actual values include zeros or near-zeros
MAPE divides errors by actual values, so any actual value at or near zero causes division by zero, making the metric undefined or infinitely large.
Holt-Winters triple exponential smoothing adds a third smoothing equation specifically for:
Answer: Seasonal factors
Holt-Winters extends Holt's method by introducing a seasonal smoothing parameter γ to track evolving seasonal indices across forecast horizons.
Which characteristic distinguishes a cyclical component from a seasonal component in time-series analysis?
Answer: Seasons have a fixed calendar period; cycles have variable duration and amplitude
Seasonal patterns repeat at known, fixed intervals (e.g., every 12 months), while cyclical swings tied to the business cycle vary in length and intensity.