The purpose of this project is development of biostatistical methods and mathematical models appropriate for the analysis of epidemiologic and experimental studies related to cancer control and prevention. Many of the statistical problems being studied under this project are derived from the consultation activities of the Section. Activities during the past year have been directed toward developing a method to reduce the bias inherent in the Robertson- Boyle solution to the nonidentifiability problem in age-period- cohort Poisson regression analysis; development of a GLIM regression method to allow unbiased estimation of the treatment effect in a randomized trial involving covariates measured with error; extension of the simple Goldie-Coldman model of tumor cell resistance to chemotherapy to include immunostimulation and time- varying birth, death, and cellular mutation; development of a computer program to statistically analyze unbalanced repeated- measures studies. This project also includes research on the development and use of mathematical models of carcinogenesis to analyze cancer studies and to predict the results of intervention strategies; the Armitage- Doll model is being used to study the effect of time patterns of carcinogenic exposure on inferences regarding their joint action; the Moolgavkar-Knudson two-stage model is being used to quantify the hypothesized effects of a woman's first full-term pregnancy upon her future breast cancer risk.