EViews - L4

Question 1

What are the different ways of estimating models with autoregressive errors?

Answer 1

• Direct way –> assuming there is a non-linear relationship and using Non-Linear Least Squares
• Intuitive way –> Using Quasi-differencing –> what was done before
• Special case –> ρ = 1

Question 2

How do you allow for an autocorrelated error in an EViews regression equation?

Answer 2

• add in AR(1) as a variable into the regession formula –> this will be our estimate of rho
• if it gets to one we actual need to differences the data rather than apply this method

Question 3

How do you calculate Stockbuilding?

Answer 3

• While inventory is a tiny proportion of GDP it is one of the most volatile
• Changes in inventory is usual an indication if we are moving to a recession or not –> big drop in demand –> dont need alot of inventory
• inventory is often negative –> thus we cannot use logs to look at any changes from one period to the next
• but when we plot the data below, our regression doesnt explain the data well –> this maybe due to the structural change in 2005 stockbuilding series, which doesnt occur in the GDP series thus the poor explanation
• Thus we need to incorporate a Dummy Variable

Question 4

What is a Dummy Variable?

Answer 4

• Using the Stockbuilding example we can add in a variable D₂₀₀₅ –> that is a value of 0 up to 2004 and a value of 1 from 2005 onwards
• we can then add this new term into the stockbuilding equation –> this now allows the intercept to be different before and after 2005 to account for this structural break and shift the constant term in the regression equation
• to enter this into EViews you create a series D2005=0 and create a sample period - smpl 2005 2017 and then set D2005 =1