# Showing papers in "The Statistician in 1993"

â€¢â€¢

6,278Â citations

â€¢â€¢

881Â citations

â€¢â€¢

473Â citations

â€¢â€¢

TL;DR: The use of Sample Surveys and Stratification and Stratified Random Sampling, and Strategies for Design-Based Analysis of Sample Survey Data, are presented.

Abstract: Tables. Boxes. Figures. Getting Files from the Wiley ftp and Internet Sites. List of Data Sites Provides on Web Site. Preface to the Fourth Edition. Part 1: Basic Concepts. 1. Use of Sample Surveys. 2. The Population and the Sample. Part 2: Major Sampling Designs and Estimation Procedures. 3. Simple Random Sampling. 4. Systematic Sampling. 5. Stratification and Stratified Random Sampling. 6. Stratified Random Sampling: Further Issues. 7. Ratio Estimation. 8. Cluster Sampling: Introduction and Overview. 9. Simple One-Stage Cluster Sampling. 10. Two-Stage Cluster Sampling: Clusters Sampled with Equal Probability. 11. Cluster Sampling in Which Clusters Are Sampled with Unequal Probability: Probability Proportional to Size Sampling. 12. Variance Estimation in Complex Sample Surveys. Part 3: Selected Topics in Sample Survey Methodology. 13. Nonresponse and Missing Data in Sample Surveys. 14. Selected Topics in Sample Design and Estimation Methodology. 15. Telephone Survey Sampling (Michael W. Link and Mansour Fahimi). 16. Constructing the Survey Weights (Paul P. Biemer and Sharon L. Christ). 17. Strategies for Design-Based Analysis of Sample Survey Data. Appendix. Answers to Selected Exercises. Index.

355Â citations

â€¢â€¢

TL;DR: This paper presents a meta-analysis of statistical errors in Nonlinear Estimates of Linear and Nonlinear Systems and their applications in Input/Output Relationships and Bilinear and Trilinear Systems.

Abstract: Linear Systems, Random Data, Spectra Zero-Memory Nonlinear Systems Bilinear and Trilinear Systems Nonlinear System Input/Output Relationships Square-Law and Cubic Nonlinear Systems Statistical Errors in Nonlinear Estimates Parallel Linear and Nonlinear Systems.

207Â citations

â€¢â€¢

TL;DR: In this paper, the value of a Bayes factor may be sensitive to the choice of priors on parameters appearing in the competing models, and the sensitivity is illustrated and discussed.

Abstract: Computational advances have facilitated application of various Bayesian methods, including the assessment of evidence in favor of a scientific theory. This is accomplished by calculating a Bayes factor. It is important to recognize, however, that the value of a Bayes factor may be sensitive to the choice of priors on parameters appearing in the competing models. This sensitivity is illustrated and discussed. The Schwarz criterion remains a crude yet useful approximation, but more accurate methods will generally require determination of priors and subjective sensitivity analysis.

â€¢â€¢

TL;DR: In this article, the use of Wand W' with grouped and with singly censored data, and of W with log-normally distributed data, is discussed, and methods are given to enable the P value of each test to be calculated under these different circumstances.

Abstract: The Shapiro-Wilk Wand Shapiro-Francia W' statistics are convenient and powerful tests of departure from normality. Modifications are described to allow the use of Wand W' with grouped and with singly censored data, and of W with log-normally distributed data. Methods are given to enable the P value of each test to be calculated under these different circumstances. It is hoped that software developers will thereby be encouraged to incorporate one or other of the tests in their statistical products.

â€¢â€¢

TL;DR: A method to help quantify beliefs in the form of a prior distribution about regression coefficients in a proportional hazards regression model is described, developed for a randomized trial comparing prophylaxes for toxoplasmosis in a population of HIV-positive individuals.

Abstract: Bayesian methods are potentially useful for the design, monitoring and analysis of clinical trials. These methods, however, require that prior information be quantified and that the methods be robust. This paper describes a method to help quantify beliefs in the form of a prior distribution about regression coefficients in a proportional hazards regression model. The method uses dynamic graphical displays of probability distributions that can be freehand adjusted. The method was developed for, and is applied to, a randomized trial comparing prophylaxes for toxoplasmosis in a population of HIV-positive individuals. Prior distributions from five AIDS experts are elicited. The experts represent a community of consumers of the results of the trial and these prior distributions can be used to try to make the monitoring and analysis of the trial robust. Bayesian approaches to clinical trials, as discussed for example by Spiegelhalter and Freedman (1988) and Freedman and Spiegelhalter (1992), require the specification of a prior distribution. In this paper, an AIDS clinical trial comparing potential prophylaxes for toxoplasmosis is used as a context to develop and implement methodology to aid in the elicitation of prior distributions. Following Kadane (1986), we require that a range of priors be identified that are representative of 'the community'. These will then be used to try to make monitoring and analysis robust. This paper focuses on the elicitation part of such an approach, and we report on the prior distributions of five individuals. Bayesian analysis and monitoring of the trial is reported in Carlin et al. (1993). First, we provide a description of the toxoplasmosis prophylaxis trial and then describe the elicitation method based on graphical input and feedback. We then summarize what we learned from eliciting the prior distributions, the results of the trial, and some of the problems in implementing a Bayesian approach.

â€¢â€¢

â€¢â€¢

TL;DR: Preface Editor's abbreviations Main dates and events in the life of R. A. Fisher.

Abstract: Preface Editor's abbreviations Main dates and events in the life of R. A. Fisher Statistical inference Statistical theory and method History of statistics Teaching of statistics History and philosophy of science Scientists and scientific research Notes on correspondents List of references to Collected Papers of R. A. Fisher Name index Subject index

â€¢â€¢

TL;DR: The hot deck multiple imputation (HMI) method as discussed by the authors is a non-parametric version of the traditional HMI, and has been used by the US Census Bureau to complete public-use databases.

Abstract: Covariate data which are missing or measured with error form the subject of a growing body of statistical literature. Parametric methods have not been widely adopted, quite possibly due to the necessity of specifying the form of a 'nuisance function' not required for complete data analysis, and the non-robustness of the methods to mis-specification. A non-parametric counterpart of multiple imputation, known as 'hot deck', was proposed by Rubin (1987) and has been used by the Census Bureau to complete public-use databases. However, inference using this method has not been possible due to the distribution theory not being available. Recently, it has been shown that the hot deck estimator has the same asymptotic distribution as the 'mean score' estimator, so that inference using hot deck is now possible. The method is intuitively appealing and easily implemented. Furthermore, it accommodates missingness which depends on outcome, which is an important generalization of many currently available methods. In this paper, the hot deck multiple imputation method is explained, its asymptotic distribution presented and its application to data analysis demonstrated by an example.

â€¢â€¢

â€¢â€¢

TL;DR: A database on the lengths of stay of patients entering a department, and their subsequent destinations, over a 16-year period, is used to examine the pattern of length of stay in ward of admission until death, discharge or transfer.

Abstract: Previous work has indicated that a two-term mixed exponential distribution gives a good fit to data on lengths of stay of patients in departments of geriatric medicine. A database on the lengths of stay of patients entering a department, and their subsequent destinations, over a 16-year period, is used to examine the pattern of length of stay in ward of admission until death, discharge or transfer. The two-term mixed exponential distribution is fitted to these data using death/discharge and transfer as the two components of the mixture in order to assess to what extent the previous success of this distribution for census data may be explained by our current longitudinal data and choice of components. The model is then extended to the more sophisticated mixed exponential and log-normal distribution which better enables us to capture the exact shape of the distribution.

â€¢â€¢

TL;DR: In this paper, the statistical power in treatment effectiveness research and design sensitivity is discussed, and a framework for estimating statistical power is presented, including dependent measures such as design, sample size, and Alpha.

Abstract: PART ONE: STATISTICAL POWER IN TREATMENT EFFECTIVENESS RESEARCH Treatment Effectiveness Research and Design Sensitivity The Statistical Power Framework Effect Size The Problematic Parameter How to Estimate Statistical Power PART TWO: USEFUL APPROACHES AND TECHNIQUES Dependent Measures Design, Sample Size, and Alpha The Independent Variable and the Role of Theory Putting It All Together

â€¢â€¢

â€¢â€¢

â€¢â€¢

Mayo Clinic

^{1}TL;DR: Computational issues are investigated by comparing the posterior distributions of model parameters obtained assuming approximate posterior normality with more precise results obtained via numerical integration and the impact of approximations on the performance of Bayesian stopping rules.

Abstract: Bayesian methods have been a subject of increasing interest among researchers engaged in the interim monitoring and final analysis of clinical trials data. In particular, Chaloner et al. (1992) have shown how prior distributions may be elicited relatively easily from a panel of experts using computer-assisted interactive graphical methods. Significant stumbling blocks remain, however, to routine implementation of Bayesian methods by practitioners. For example, heavy computational burdens have historically precluded real-time evaluation of stopping rules. In addition, the resulting statistics may not be amenable to simple monitoring displays, since the resulting posterior distributions for important model parameters need not be symmetric, unimodal or based on a low-dimensional sufficient statistic. In this paper we investigate these computational issues by comparing the posterior distributions of model parameters obtained assuming approximate posterior normality with more precise results obtained via numerical integration. We also investigate the impact of approximations on the performance of Bayesian stopping rules. Where the normal approximation is inappropriate, the Bayesian methodology still allows for inference and simple monitoring displays based on posterior probabilities. We illustrate the methodol- ogy with a numerical example featuring prior distributions elicited from five AIDS experts and data from a recently completed toxoplasmic encephalitis prophylaxis trial (Jacobson et al., 1992).

â€¢â€¢

TL;DR: In this article, a statistical analysis of the increasing inflow of data with which astronomers are confronted from different modern facilities is presented, stemming from a meeting held in September 1989 at the International Stellar Data Centre in Strasbourg.

Abstract: A statistical analysis of the increasing inflow of data with which astronomers are confronted from different modern facilities. This book stems from a meeting held in September 1989 at the International Stellar Data Centre in Strasbourg.

â€¢â€¢

â€¢â€¢

TL;DR: In this article, a Bayesian approach is used to estimate the sample size required to estimate a parameter, which is then determined by requiring the posterior distribution to have particular properties, such as the desired properties of the posterior distributions for the parameter of interest.

Abstract: The design of scientific studies is a significant part of statistics, an important aspect of which is considering the number of experimental units required. The formulae used in classical statistics assume that the sample size should be determined by controlling the variance of the parameters of interest or the type I and type II error rates for testing hypotheses. A Bayesian approach is based on considering prior distributions and the desired properties of the posterior distributions. Four methods for calculating sample sizes using a Bayesian philosophy are presented. As these methods are particularly relevant when there are precise but conflicting prior distributions, a study of professional's beliefs about indicators of child sexual abuse is used to illustrate the methods. All applied statisticians are aware of the need for careful design of scientific studies. Although, unfortunately, advice is often requested only about the number of experimental subjects needed, the calculation of sample sizes is important. Classical methods require the scientists to know or estimate the value and the variance of the parameters of interest, and specify the type I error rate, in order to use formulae for the number of units necessary to obtain a given precision. The type II error rate must also be specified if a significance test comparing parameters is considered. Instead of requiring a specified value and a variance, a Bayesian approach attempts to quantify the scientist's knowledge by defining a prior distribution for the parameter of interest. The sample size required to estimate a parameter is then determined by requiring the posterior distribution to have particular properties. Comparison of different possible values of a parameter is dealt with by specifying relationships between the posterior distributions for the parameter, given different prior distributions. We consider four cases with respect to the probability of success in a Bernoulli trial: first, the implications of the classical sample size calculation given an individual's prior distribution; second, requiring the posterior credible interval to include a specific value; third, requiring the posterior credible interval for an individual to include a parameter with an alternative prior distribution; and, finally, requiring the posterior credible intervals for two extreme prior distributions to touch. Although there is now considerable interest in using Bayesian methods in the analysis of health-related studies, there has been relatively little discussion of the use of Bayesian ideas in the design of studies. Herson (1979) considers the sensitivity of a Bayesian stopping rule for phase II clinical trials to different prior distributions, but assumes that one prior distribution will be specified. McPherson (1982) also considers the effects of different prior distributions, in this case the effect of the prior distribution of the treatment difference on the choice of the number of interim analyses in clinical trials. He comments in passing that the prior distribution could be obtained by averaging over experts' priors. Freedman and Spiegelhalter (1983) describe a study which uses McPherson's methods. They suggest either an average prior distribution or the use of two or three consensus prior distributions, rather than attempting to use divergent prior distributions. Spiegelhalter and Freedman (1986) use a consensus prior to calculate the probabilities of reaching the correct or