The objective is to further develop practical statistical methods necessary for data involving times to response. Such data are central to many areas of health-related research, including clinical trials, laboratory experimens eith animals, and epidemiological cohort studies. It is well known that ordinary regression analysis and analysis-of-variance methods are often unsuitable for such data, and there has been much development in recent years of more suitable statistical methods. The investigator has done work of a moderately general nature in this area under the currently-funded grant. What is proposed here is to build on this in more specific and applied ways. Focus will be substantially on analysis of cancer incidence related to environmental exposures, using epidemiological cohort data. These large and complex data sets present several difficulties in applying recently developed methods for response-time analysis. Some of the specific issues to be addressed relate to issues of: (i) feasibility of using regression-type models for excess risk expressed in an additive manner, in relation to background risk, as opposed to more commonly used multiplicative models, (ii) methods for comparison of adequacy of such additive and multiplicative models, (iii) comparison of methods using parametric and nonparametric models for background risk, and (iv) use of information external to the study population regarding background risks.
