18
additional virtue of the Wordscore application is that it also computes standard errors for each
estimate, which facilitates formal statistical tests. For our purpose here, we simply note that the
confidence intervals constructed around these estimate would involve virtually no overlap, so one
can be confident of the relative ordering of these cases.
Not less important, we want to know just where these policies are located in dimensional
space, especially with respect to our reference texts. To illustrate their respective locations, we
graph these point estimates alongside the values of the reference texts and present these data in
Figure 1.
10
10
In this and in subsequent figures, we add a uniform value (in this histogram, we add 2) so that each estimate
takes on a positive value. We do this simply to facilitate the graphical representation of the data.
Figure 1. Wordscores for Free Exercise Cases
0
0.5
1
1.5
2
2.5
3
3.5
4
O'Lone v. Shabazz (1987)
Goldman v. Weinberger (1986)
Heffron v. ISKCON (1981)
Employment Div v. Smith (1990)
U.S. v. Lee (1982)
Convention