Death Penalty and Race
Lawyers collected data on convicted black murderers in the state of Georgia to see whether convicted black murderers whose victim was white were more likely to receive the death penalty than those whose victim was black, after accounting for aggravation level of the murder. They categorized murders into 6 progressively more serious types. Category 1 comprises barroom brawls, liquor-induced arguments lovers' quarrels, and similar crimes. Category 6 includes the most vicious, cruel, cold=blooded, unprovoked crimes.
case1902
A data frame with 12 observations on the following 4 variables.
Aggravation
the aggravation level of the crime, a
factor with levels "1"
, "2"
, "3"
, "4"
,
"5"
and "6"
Victim
a factor indicating race of murder victim, with
levels "White"
and "Black"
Death
number in the aggravation and victim category who received the death penalty
Nodeath
number in the aggravation and victim category who did not receive the death penalty
Ramsey, F.L. and Schafer, D.W. (2002). The Statistical Sleuth: A Course in Methods of Data Analysis (2nd ed), Duxbury.
Woodworth, G.C. (1989). Statistics and the Death Penalty, Stats 2: 9–12.
str(case1902) # Add smidgeon to denominator because of zeros empiricalodds <- with(case1902, Death/(Nodeath + .5)) plot(empiricalodds ~ as.numeric(Aggravation), case1902, log="y", pch=ifelse(Victim=="White", 21, 19), xlab="Aggravation Level of the Murder", ylab="Odds of Death Penalty") legend(3.8,.02,legend=c("White Victim Murderers","Black Victim Murderers"),pch=c(21,19)) fitbig <- glm(cbind(Death,Nodeath) ~ Aggravation*Victim, case1902, family=binomial) # No evidence of overdispersion; no statistically significant evidence # of interactive effect anova(fitbig, test="Chisq") fitlinear <- glm(cbind(Death,Nodeath) ~ Aggravation + Victim, case1902, family=binomial) summary(fitlinear) # Mantel Haenszel Test, as an alternative table1902 <- with(case1902, rbind(Death,Nodeath)) dim(table1902) <- c(2,2,6) mantelhaen.test(table1902)
Please choose more modern alternatives, such as Google Chrome or Mozilla Firefox.