During the last twenty years statistical data and inference have played an increasing role in establishing basic facts used in determining public policy, e.g., estimating the prevalence of HIV infection, evaluating the risk of cancer from occupational or environmental exposures, assessing changes in the income distribution and in the legal environmental setting, especially in cases concerning employment discrimination and school segregation. The purpose of the project is to study statistical problems arising in these contexts and to develop appropriate probability models and statistical methodology for their solution. Robust methods which lead to statistically valid inferences when several scientifically plausible models may underlie the data or when a few observations may be erroneous are emphasized. Different approaches for assessing the potential effect of an omitted factor and measurement errors on inferences drawn from otherwise valid studies will be developed. The completion of this project should establish the applicability of group testing to the identification, quality control and confidential estimation of the prevalence of a trait (such as the HIV virus) in a population or workforce. The successful completion of the second task of the project, namely, the development of methods for assessing the statistical bias resulting from omitted factors and misclassification in surveys, could have a very important impact in the legal and public policy setting as courts have claimed that even one error can justify criticism of a statistical study. The principal investigator has already made significant progress in unifying approaches used in various areas of application and in assessing the seriousness of the possible error on the conclusions drawn from a study.
