Accumulate-Toward-the-Mode Approach to Confidence Intervals and Hypothesis Testing With Applications to Binomially Distributed Data

摘要

Analyses of binomially distributed data usually depend on the normal approximation to the binomial or, for small sample sizes, the I binomial cumulative distribution function. The first approach is only good for an unspecified 'sufficiently large' sample size and does not reflect the asymmetry of the binomial distribution for probabilities other than 0.5. The second approach leads to suboptimal designs. .Neither approach extends easily to handle multivariate binomial data. This paper describes a new accumulate toward the mode approach to hypothesis tests and confidence intervals and the application of this approach to binomially distributed data. The extension of this approach ' to multivariate data is also covered.