Which method is listed for horizontal secular trend with seasonality?

Which method is listed for horizontal secular trend with seasonality?

• A. Seasonal Exponential Smoothing
• B. Multiple Regression
• C. Naive
• D. Moving Average

Answer: (B)

Modeling a horizontal secular trend with seasonality is best handled by a setup that keeps a constant average level while explicitly capturing seasonal effects. Multiple regression does exactly this by including a constant term (to represent the constant level) and seasonal dummy variables (to capture the systematic ups and downs for each season or month). With this structure, you estimate one average level and separate seasonal shifts, so the fitted values reflect a flat trend over time with the seasonal pattern added on top. For example, you’d specify y_t = β0 + β1·D1_t + β2·D2_t + ... + ε_t, where the D’s are seasonal indicators. One season is typically omitted as the baseline, so the intercept β0 represents the average level across seasons, and the other coefficients adjust that level for each season. This approach directly targets a constant level while allowing season-specific effects to vary, which is exactly what a horizontal trend plus seasonality requires. Naive relies only on the last observation and misses the explicit seasonal structure. Moving average smooths data without separately estimating a seasonal effect or a constant level. Seasonal exponential smoothing can model seasonality, but it implicitly combines level and seasonal components and is less flexible when you want to explicitly test or enforce a horizontal trend with seasonality via regression coefficients.

Modeling a horizontal secular trend with seasonality is best handled by a setup that keeps a constant average level while explicitly capturing seasonal effects. Multiple regression does exactly this by including a constant term (to represent the constant level) and seasonal dummy variables (to capture the systematic ups and downs for each season or month). With this structure, you estimate one average level and separate seasonal shifts, so the fitted values reflect a flat trend over time with the seasonal pattern added on top.

For example, you’d specify y_t = β0 + β1·D1_t + β2·D2_t + ... + ε_t, where the D’s are seasonal indicators. One season is typically omitted as the baseline, so the intercept β0 represents the average level across seasons, and the other coefficients adjust that level for each season. This approach directly targets a constant level while allowing season-specific effects to vary, which is exactly what a horizontal trend plus seasonality requires.

Naive relies only on the last observation and misses the explicit seasonal structure. Moving average smooths data without separately estimating a seasonal effect or a constant level. Seasonal exponential smoothing can model seasonality, but it implicitly combines level and seasonal components and is less flexible when you want to explicitly test or enforce a horizontal trend with seasonality via regression coefficients.