ABSTRACT
This chapter considers the methodologically significant inference problem for one binomial proportion. It begins by discussing the Wald interval, which is the most common and often uncritically used method and points out its lack of reliability due to the erratic nature of the coverage probability. The chapter introduces wide-ranging alternatives to the Wald interval. The methodologies discussed in this chapter can be broadly classified into three groups, namely, asymptotic, Bayesian and exact methods. The first part of the chapter discusses various asymptotic methods, the second part covers Bayesian intervals and the last part introduces a range of exact methods. As measured by the nominal coverage probability and expected lengths, the performances of these intervals are discussed in the subsequent section. Finally, the chapter concludes with discussions and recommendations. All methods covered in the chapter include detailed SAS codes and are illustrated with examples from clinical trials.
