Previous: How many observations do

Up: Frequently Asked Questions

Next: What if I find

Has this method been used in Court?

A federal district court judge declared that A Solution to the Ecological Inference Problem ``is the best method currently available to measure racial bloc voting'' [i.e., whether blacks and whites vote differently] (Mallory, et al., v. State of Ohio, et al., USDC Southern District Court of Ohio, case no. C-2-95-381.) This decision was upheld on appeal before the three judges of the Circuit Court, and similar language was used to describe the method (in Mallory, et al. v. State of Ohio, et al., United States Court of Appeals, Sixth Circuit, No. 97-4425, 4/13/99). It was also accepted by the court in Bill Jones vs. Ca Democratic Party and in a variety of others. In each case, EI was unchallenged by experts on the opposing side. Since EI starts with Goodman's regression, which had previously been declared as the appropriate method by the Supreme Court, and adds information known to be true (i.e., the bounds), EI has been relatively uncontroversial in these situations.

I have not collected all uses of EI by the courts, but a recent eloquent statement was given in Judge Gruender's concurring opinion in Bone Shirt v. Hazeltine (U.S. Court of Appeals, eighth circuit, No. 05-4010, on appeal from District Court of South Dakota, filed 22 August 2006, p.20), where he writes ``I find it difficult to rely upon [bivariate ecological regression analysis (BERA), i.e., Goodman's regression,] a statistical method that incorporates an admittedly erroneous equation that yields a result with an error of unknown quantity and effect. Daubert specifically requires a district court to consider `the known or potential rate of error' of the scientific method utilized by testifying experts. 509 U.S. at 594. It is difficult to weigh this factor in Daubert's analysis if `the effect of that error is unknown.' 336 F. Supp. 2d at 1002. Nor am I persuaded that previous acceptance of BERA results permanently decides the matter. Science evolves, and scientific methods that were once considered unassailable truths have been discarded over time. Unreliable testimony based upon those outdated theories and methods must be discarded as well, lest scientific stare decisis ensure that such theories survive only in court. It may be that the validity of some scientific methods need not be constantly reestablished in case after case, but a statistical method that contains an apparently unquantified error would not be one of them. If Cole's BERA-based testimony were the only basis for the district court's opinion, I would likely hold that its conclusions were unsupported by competent evidence.''