Chapter 10 (Dynamic regression models)

Question 1

Dynamic regression models

Answer 1

allow the errors from a regression to contain autocorrelation

shown by ‘nt’

Question 2

Deterministic trend

Answer 2

yt = B0 + B1t + nt

nt is ARMA (no differencing)

Question 3

Stochastic trend

Answer 3

yt = B0 + B1t + nt

nt is ARIMA (with differencing)

Question 4

Problems with OLS

Answer 4

1. Estimated coefficients are inconsistent
2. Standard errors are incorrectly calculated
3. Statistical inference is invalid
4. AIC can be misleading

Question 5

Why use ARIMA structure on regression model?

Answer 5

1. corrects the problems
2. short term forecasts are more accurate

Question 6

Dynamic Harmonic Regression - Fourier terms with ARIMA errors

Answer 6

Benefits:
- captures seasonality with different frequencies
- controls smoothness of seasonal patterns by K
- short-term dynamics are easily handled

Con:
- seasonality is assumed to be fixed