Basic and Applied Studies of Statistical Inference

近年来，统计数据和推断在确定用于各种领域的公共决策的基本事实方面发挥了重要作用，如估计失业率、确定职业接触有毒化学品所致癌症的风险以及评估艾滋病抗体或药物使用筛查测试的准确性。在法律领域，使用统计方法检查雇用和晋升数据，以寻找歧视的证据，并评估纳税评估的公正性。事实上，最初为分析医疗数据而开发的方法已被用于分析雇用和晋升信息--当前项目的主要调查人员一直在积极研究这一领域以及上文所述的一些问题。特别是，在艾滋病危机的推动下，调查人员将使用建议奖下提供的资金，进一步研究疾病流行率的估计值和筛查测试在低流行率(低于100%)人群中的预测价值。他们计划采用最初为经济数据设计的保密技术，以建议获得可靠的艾滋病流行估计的程序，同时仍保留接受测试的个人的匿名性。根据推荐的奖励，调查人员还计划调整用于评估医疗筛查测试的方法，以评估人口普查局主要失业调查中的明显反应偏差，并研究其他测量问题，这些问题需要开发统计程序，即使在所有通常的假设不被数据满足的情况下也能产生有效的结果。这项研究预计将对统计理论作出贡献，除了它在向最近认识到的、肯定会在未来几年伴随我们的问题提供应用方面的及时性之外。

In recent years, statistical data and inference have played an important role in establishing the basic facts used for public policymaking in a variety of areas, such as estimating the unemployment rate, determining the risk of cancer from occupational exposure to toxic chemicals, and the evaluation of the accuracy of screening tests for AIDS antibodies or for drug use. In the legal area, statistical methods are used to examine hiring and promotion data for evidence of discrimination, and also to appraise the fairness of tax assessments. Indeed, methods originally developed for the analysis of medical data have been used to analyze hiring and promotion information -- and the principal investigators of the current project have been active in researching problems arising in this area as well as some of those indicated in the paragraph above. In particular, motivated by the AIDS crisis, the investigators will use funds to be provided under the recommended award to pursue further research into estimators of the prevalence of a disease and the predictive value of screening tests in populations with low prevalence (less than one in one hundred). They plan to adapt confidentiality techniques originally designed for economic data to suggest procedures for obtaining a reliable estimate of AIDS prevalence while still preserving the anonymity of the individuals tested. Under the recommended award the investigators also plan to adapt methods used to evaluate medical screening tests to evaluate an apparent response bias in the Census Bureau's major unemployment survey, and to study other measurement problems requiring the development of statistical procedures which produce valid results even when all the usual assumptions are not satisfied by the data. The research is expected to make contributions to statistical theory in addition to its timeliness in proferring applications to recently recognized problems that are certain to be with us for some years to come.