In your own words, describe the 3-steps in the process of improving a short-term
ID: 3857959 • Letter: I
Question
In your own words, describe the 3-steps in the process of improving a short-term forecast by using an AR(1) model of the residuals.
Question 1 options:
Question 2
Autoregressive models are based on the assumption that
Question 2 options:
The data has only a linear trend.
A moving average model can be used to completely explain variation in the data.
Computer software can automatically do regression when given appropriate data.
Chronological observations in a time series are expected to be related to one another (next month’s value is related to this month’s value which is related to the previous month’s value and so on).
Question 3
To calculate the lag-2 autocorrelation for a time series you should
Question 3 options:
Compute the correlation of the time series with the value of the time series squared.
Compute the correlation of each observation in the time series with the observation from 2 periods before that observation.
Compute the correlation of the time series with a series created by subtracting 2 from each value.
Compute the correlation of the time series with the value of the times series times 2.
Question 4
If gpd.ts is defines as a time series containing monthly values for gross domestic product the R forecast package command Acf(gdp.ts, lag.max = 12, main = “”) will
Question 4 options:
Return the value of which lag (1 to 12) will have the largest positive autocorrelation
Tell you which lag value for your AR(#) model will maximize the likelihood function
Create 12-step ahead forecasts using an AR(1) for the gdp.ts series
Create a graph showing the autocorrelations for this time series for lag-1, lag-2, … , lag-12
Question 5
In a quarterly series, strong positive autocorrelation at lags 4, 8, and 12 would likely indicate
Question 5 options:
That the series is non-stationary
A linear trend
Seasonality
That the series is a random walk
Question 6
Examining the autocorrelations from the residuals of a forecasting model can be useful because
Question 6 options:
We can determine if we have adequately captured seasonality in our model
Strong autocorrelations in the residuals indicate the model is over fitted
Strong autocorrelations in the residuals indicate a model that fits the data well
We should see a positive negative positive negative repeating pattern in the residuals
Question 7
ARIMA models require the forecaster to choose 3 parameters, p, d and q. These represent
Question 7 options:
The acceptable p-value for the results (usually 0.05), the value of delta (expected change from one period to the next), and the number of periods to be forecast
Price, demand, quantity sold
Probability of the series increasing, the probability of the series decreasing, the quantity of increase or decrease
The time series autocorrelation lag, the number of times the series is differenced, the forecast error autocorrelation lag
Question 8
To test whether a series is non-stationary (that is a random walk) you can
Question 8 options:
Fit and ARIMA(1,2,3) model and examine the autocorrelations of the residuals (if they are all zero, the series is a random walk)
Compare an AR(1) and MA(1) model; if neither is more accurate, then the series is a random walk (not predicable)
Fit an AR(1) model and test the hypothesis that the slope coefficient on the lagged value is equal to 0
Fit an AR(1) model and test the hypothesis that the slope coefficient on the lagged value is equal to 1
Question 9
It we wanted to incorporate the effects of the September 11, 2001 terrorist attacks in our forecasting model of air passengers, we could
Question 9 options:
Only use the pre-9/11 data in building our model
Use a lag-12 model to determine how much lower passenger demand was in October 2001 compared to October 2000
Include a dummy variable equal to 1 for observations post-9/11 and equal to 0 pre-9/11
Apply an ARIMA(9, 11, 1) model
Explanation / Answer
Question 2 : A moving average model can be used to completely explain variation in the data.
question 3: Compute the correlation of the time series with the value of the time series squared.
question 4: Create a graph showing the autocorrelations for this time series for lag-1, lag-2, … , lag-12
question 5: Seasonality
question 6: We should see a positive negative positive negative repeating pattern in the residuals
question 7: The acceptable p-value for the results (usually 0.05), the value of delta (expected change from one period to the next), and the number of periods to be forecast
question 8: Compare an AR(1) and MA(1) model; if neither is more accurate, then the series is a random walk (not predicable)
quesgtion 9: Use a lag-12 model to determine how much lower passenger demand was in October 2001 compared to October 2000