Consulting for the City of New York, statisticians at Analysis & Inference designed an experiment that helped settle a multi-million dollar dispute with a vendor. The dispute was over the performance of new garbage trucks purchased from a vehicle manufacturer. Did new garbage trucks fulfill the contract specification that the trucks load a minimum full-load of 12.5 tons of household refuse? Although the average truck load was determined to be over 12.5 tons, some loads fell below this value. Using a classical design, the “Latin Square,” the statistical analysis clarified that the below-spec loads were not due to the capacities of the trucks themselves but to environmental factors. Modern “Bayesian” analysis showed how the evidence would be persuasive to both the City and the manufacturer, despite their differing prior views. Access PDF of article here.
In a paper to appear in Observational Studies in February 2018, statisticians William Fairley and William Huber at Analysis & Inference review how courts have treated statistical evidence of causality in prima facie proof of disparate impact discrimination. Noting differences in statistical and legal criteria for proof, they observe that in both realms providing a “tenable alternative hypothesis” may not be required but is more persuasive.