What's new

P1.T2.700. Seasonality in time series analysis (Diebold)

Nicole Seaman

Director of FRM Operations
Staff member
Subscriber
Learning objective: Describe the sources of seasonality and how to deal with it in time series analysis.

Questions

700.1. Which of the following time series is MOST LIKELY to contain a seasonal pattern?

a. Price of solar panels
b. Employment participation rate
c. Climate data data recorded from a weather station once per year
d. Return on average assets (ROA) for the large commercial bank sector

700.2. In order to forecast housing starts, your colleague Brett is going to use the following seasonal model that employs a regression on seasonal dummies:

a. This model do not contain a trend
b. If the model includes twelve monthly seasons (January, February, ..., December) such that s = 12, then seasonal factor γ(5) is probably greater then either γ(1) or γ (12)
c. If the model includes four seasons (spring, summer, fall and winter) such that s = 4, then he should include four (4) seasonal dummy variables plus an intercept
d. If this is a quarterly model such that y(t) = + D2 + D3 + 9.0*D4, and the standard error of the y(4) = 9.0 coefficient is 15.0, then we can infer that average housing starts in the 4th quarter is not statistically different than average housing starts in the first quarter

700.3. In regard to modeling and forecasting seasonality, each of the following is true EXCEPT which is not accurate?

a. A seasonal time series is, by definition, covariance stationary
b. Trading-day variation is a type of seasonality that refers to the fact that different months contain different numbers of trading days
c. A key technique for modeling seasonality is regression on seasonal dummy variables; dummy variables assume a value of zero or one.
d. Log transformation is useful in both trend models and seasonal models, but for different reasons; in a seasonal model, log transformation can stabilize seasonal patterns whose variance is growing over time