Statistical models for proportions and probabilities (Q354274): Difference between revisions

Most elementary statistics books discuss methods for making inferences from data about one or more probabilities and proportions. Since some of the recommended so-called `large-samples' rules in textbooks have been found to be inappropriate, this monograph endeavors to give an up-to-date review of recent research on this topic and presents both exact methods and helpful approximations. In this frame, the present monograph consists of five chapters. In the sequel, the contents of each one is briefly presented. Chapter 1 (Single proportions) initially looks at the difference between the two concepts, proportions and probabilities. Afterwards, it focuses on the problem of estimating a population proportion using random sampling with or without replacement or inverse sampling. Since an important model involving a probability is the binomial distribution, in Chapter 2 (Single Probabilities) this distribution and some of its properties, including maximum likelihood estimation, are discussed. Moreover, exact and approximate confidence intervals for the single probability are provided. Chapter 3 (Several proportions or probabilities) focuses on exact and large sample inference for comparing two proportions or two probabilities from the same or different populations, as well as on some related quantities such as the relative risk and the odds ratio. Theory is also presented for the multi-hypergeometric and multinomial distributions which are involved. For instance, maximum likelihood estimation, asymptotic normality theory and simultaneous confidence intervals are given for the multinomial distribution. The chapter closes with some simple applications to animal populations and with an appendix. In the appendix, the theory of the delta method is briefly discussed and then applied to two different cases. In Chapter 4 (Multivariate hypothesis tests) the asymptotic equivalence for testing hypotheses about the multinomial distribution of the likelihood-ratio test, the Wald test, the score test and Pearson's goodness of fit tests is established. In this frame, particular attention is given to contingency tables, especially 2x2 tables and matched data. Chapter 5 (Logarithmic models) provides some general background for two logarithmic models for discrete data, namely the log-linear and logistics models. In this context, these models are considered for handling contingency tables, multinomial distributions and binomial data. An application of the log-linear model to epidemiology is discussed. This short book through its lucid description is able to be useful for those teaching statistics at a university level and for those involved in statistical consulting. At the end of each of the chapters references are provided and this will help a reader wishing to pursue this area further.