Consider the data found in the file, "gas10." You are interested in the effect that a gas tax has on petroleum consumption. The data is cross-sectional and the variables are as follows:
PCON = petroleum consumption (millions of BTU's)
REG = motor vehicle registrations
POP = population
TAX = gas tax (cents per gallon)
a) Run an appropriate regression to test the relationship between the gas tax and petroleum consumption.
b) Using REG as your proportionality factor, conduct appropriate tests to determine whether there is heteroskedasticity among the error terms of your equation. Describe what you have done, and interpret your results. If there is heteroskedasticity, how does it affect your regression results?
c) Using REG as your proportionality factor, rerun your regression applying weighted least squares. Interpret your results.
Since petroleum consumption is highly related to population, we should include POP in our regression. (Note that "motor vehicle registrations" is also highly related to population, we shouldn't include it in order to avoid multicolleanearity)
The regression should be :
PCON = a ...
The solution explains the concept of Heteroskedasticity and also runs an appropriate regression model. The regression is run using SPSS. The solution is well explained and easy to read. The output (as it is from a statistical software) may not be easy to ready and understand by some students. However, the solution provided goes into some detail about the solution.
Overall, a mediocre solution to the problem provided.
Heteroskedasticity Residual Plots
I am attaching three different residual plots. I am trying to determine whether heteroskedasticity is present. I don't believe I see any pattern indicative of heteroskedasticity (nonconstant variance). I know it can be difficult to detect - do you see any indication of this?View Full Posting Details