STA442/1008 Assignment 4

Quiz on Friday Oct. 21st in tutorial

This assignment is based upon Chapter 3 and associated lecture material.

Remember, the idea here is that you can test the association between two categorical variables controlling for a third variable, by producing separate tables (and associated chisquare statistics) for each value of the third variable. Then you add chisquare values and add degrees of freedom.

The (non-random) sample here consists of prisoners in U.S. Jails (in Florida, I believe) who were convicted of first degree murder. The prisoners are either Black or White, the victims were either Black or White, and the prisoner either got the death penalty or not. If you are interested in more detail, the reference is Am. Soc. Review 1981, 46, 918-927. Of course you do not need to go to the library to do this assignment.

The data are available in the file deathpen.dat. The variables are

  1. Race of Prisoner (1=White 2=Black)
  2. Race of Victim (1=White 2=Black)
  3. Death Penalty (1=Yes 2=No)

Please do the following.

  1. Examine all three 2-way tables, and do chisquare tests. What would you conclude?
  2. Test the association between Race of Prisoner and Death Penalty, controlling for Race of Victim. Give the value of the test statistic (a number), the degrees of freedom and the p-value. What, if anything, do you conclude about race and the death penalty?
  3. Test the association between Race of Victim and Death Penalty, controlling for Race of Prisoner. Give the value of the test statistic (a number), the degrees of freedom and the p-value. What, if anything, do you conclude about race and the death penalty?

Note: SAS warns us about a cell with expected frequency less than 5. In fact, it's also less than one, but don't worry. That sub-table is contributing a small value to the chisquare statistic, not a large one, so it is not inflating Type I error -- and that's what we care about, mostly. So, please don't worry about the warning. No trap is being laid for you here.