Regression methods for count data with extra-binomial variation

Statistical methods to analyze RERF chromosome aberration data are discussed. These data involve strong extra-binomial variation and appropriate methods to examine dose (radiation) response relationships are required. The use of a correlated probit regression model has potential for such a purpose and, in addition, possesses a natural interpretation. Methods for parameter estimation in this model are under development, as described herein. An illustration of the use of the correlated probit model is given. Other alternative procedures are also briefly discussed. 22 references, 3 tables