Home > Statistics Every Writer Should Know > The Stats Board > Discusssion
Legal Stats
Message posted by Kari (via 63.253.198.219) on December 8, 2001 at 7:20 PM (ET)
Here is a problem posed in one of my law school classes--could anyone help me out? Let me know what you think:
"The following is an old but real case. Evaluate statistically (i.e. the p-value and conclusion)not merely expressing an opinion.
An Alambama court case (Swain v. Alambama) involved a claim of discrimination against blacks in the selection of grand juries. The proportion of blacks in the eligible population was 25%, but only 177 blacks among 1050 were called for jury duty. The court did not find convincing evidence that there was discrimination. Evaluate the following statement that was written by a law professor:
'That trivial differences can appear statistically significant only underscores the admonition that the p-value should not be considered in a vacuum. THe courts are not likely to lose sight of the question of practical significance and to shut their eyes to the possibility that the degree of discrimination [in this case] is itself [minimal].'"
Please, anyone, let me know your views on this, and how you used the p-value to reach your conclusions. Thank you!
READERS RESPOND:
(In chronological order. Most recent at the bottom.)
Re: Legal Stats
Message posted by mbd (via 152.163.195.177) on December 19, 2001 at 7:52 PM (ET)
As I recall, p-values are used in statistical inference, when trying to answer questions about the unknown parameters of a population. But in this case, the parameter of interest is known (pi = .25), so a better question might be: "Is the sample proportion reasonable on the basis of the population proportion?"One way to answer this is to compute p-chart limits, mean plus or minus three standard deviations, where the mean is .25 and the standard deviation is sqrt[.25*(1-.25)/1050]. This assumes that the population is fairly large relative to the sample (n/N<.05).
I get about .21 for the lower limit and .29 for the upper limit. The sample proportion is about .169, which is outside the limits, so I might be tempted to conclude that something unusual is going on, maybe discrimination.
mbd
Your $5 contribution helps cover part the $500 annual cost of keeping this site online.