Homework Assignment #7: Multivariate Regression Analysis
Upon returning from spring break, you find a note on your desk from Mr. John Daniels,
the Director of Research for the think tank, Spirits in a Safer Society. Daniels’ note
conveys his hope that you would be willing to conduct multivariate research on behalf of
the think tank. Mr. Daniels thinks that American society has become drunk on the belief
that society’s alcohol consumption explains the incidence of vehicle fatalities. Mr.
Daniels confesses to being “a little old fashioned”, still holding to the belief that “drivers
cause accidents”, not alcohol.
Still, Mr. Daniels is a reasonable man and wishes to fully appreciate the relationship
between liquor and vehicle fatalities. He asks that you use the Auto Fatality Dataset
which he has found on the Stock and Watson website under Datasets for Replicating
Empirical Results (He wishes that you use as many of the following explanatory
variables as you deem appropriate in order to explain the vehicle fatality rate (mrall):
• spirits consumption (spircons)
• the amount of tax on a case of beer (beertax)
• the percentage of drivers aged 15-24 (yngdrv)
• the percentage of the state’s population that is Southern Baptist (sobapt)
• the average number of miles per driver (vmiles)
• the percentage of the state’s population that live in dry counties (dry)
• the alcohol-involved vehicle fatality rate (mraidall)
He wants to know if his old fashioned belief is simply out of fashion (but not inaccurate)
or if the teetotalers and surgeons general are right, that alcohol consumption really is a
significant determinant in the variance of vehicle fatalities. He requests that you
communicate the substance of your analysis in memo format. He wishes that you make
sure to explain the logic you used when establishing your model.
In addition to your analysis, Mr. Daniels cherishes his intellectual independence and,
consequently, wants to be able interpret the data on his own. He requests, therefore, that
you address whether he should be on the lookout for (1) partial effects, (2) the principle
risks of model building, (3) multicollinearity and (4) interaction effects. Finally, some
aides encouraged Mr. Daniels not to “waste his time with statistical hacks” and instead
engage in step-wise regression. He wants to know if he was right to ignore these aides.
Robert will be faxing your report to Mr. Daniel’s home in Tennessee at 5:05pm Friday
afternoon. Please put your work in Robert’s box by 5pm.
docsity.com
