ABSTRACT
This chapter contains material on the design of experiments when each observation is assumed to come from the Bernoulli or binomial distributions. The three commonly used link functions, the logit, probit and complementary log-log, will each be considered. Examples will be given of constructing designs for the cases of m = 1, m = 2 and m > 2 explanatory variables. It will be shown how to obtain a locally D-optimal design for a specified parameter set if the locally D-optimal design for a related parameter set is already known. Obtaining exact designs (see Sub-section 3.3.1) will be discussed. Obtaining a design when the total number of observations is small will be considered. Lastly, there will be brief discussion on obtaining a design that is optimal when there is uncertainty about the appropriate link function to use, or which predictor variables to use, or the form of the linear predictor.
