Showing papers in "Journal of The Royal Statistical Society Series A-statistics in Society in 1990"

Multi-Armed Bandit Allocation Indices.

Abstract: 3. Multi‐armed Bandit Allocation Indices. By J. C. Gittins. ISBN 0 471 92059 2. Wiley, Chichester, 1989. xii + 252pp. £29.95.

A Point Process Modelling Approach to Raised Incidence of a Rare Phenomenon in the Vicinity of a Prespecified Point

Abstract: Motivated by the current debate on possible raised incidence of certain types of cancers near nuclear installations, this paper develops a methodology for fitting a class of inhomogeneous Poisson point process models to data consisting of the locations of all occurrences of some phenomenon of interest within a designated planar region. The model is based on a multiplicative decomposition of the intensity function, with separate terms to describe natural spatial variation in intensity and possible raised incidence around a prespecified point. A nonparametric kernel smoothing approach, based on data from a related phenomenon, is used to describe natural spatial variation, while a parametric maximum likelihood approach is used to describe raised incidence near the prespecified point. The methodology is applied to data on the spatial distribution of cancers of the larynx and of the lung in the Chorley‐Ribble area of Lancashire, England.

The Use of Performance Indicators in the Public Sector

Abstract: Performance indicators have become ubiquitous in the public sector. In this paper, recent developments are assessed in relation to work in the private sector, where public reporting of activity has a much longer history. It is shown that lines of accountability are far more complex in the public sector. A history of initiatives to 1989 in the UK is followed by a brief survey of the analytic techniques available to interpret performance indicators. The paper ends with an attempt to trace the purposes underlying the various initiatives described and warns of the possible inefficiencies that they might induce.

Record Linkage: Statistical Models for Matching Computer Records

Abstract: We wish to measure the evidence that a pair of records relates to the same rather than different individuals. The paper emphasizes statistical models which can be fitted to a file of record pairs known to be correctly matched and then used to estimate likelihood ratios. A number of models are developed and applied to U.K. immigration statistics. The combination of likelihood ratios for possibly correlated record fields is discussed. A series of comments on the paper is also included as well as a reply to those comments by the author (pp. 312-20). (EXCERPT)

Class Inequalities in Education in the Twentieth Century

Abstract: SUMMARY This paper reanalyses data from four nationally representative sample surveys, conducted in 1949, 1972, 1983 and 1987, to explore trends in class inequalities in education. Log-linear models are used to test whether the relationship between father's class and respondent's education varies significantly by birth cohort. The conclusion is that the various educational reforms that have been attempted over the past 50 years in Britain have had little impact on class inequalities in education. Educational reform has had many objectives-the socialization of the lower classes into civic responsibility, the increase of economic efficiency, the extension of meritoc- racy and the reduction of inequalities between the classes. It is the last of these which has been most amenable to empirical research and which has been a major focus of sociological endeavour. In Britain we now have a series of major national surveys from which we can chart the history of class inequalities in education over the course of the present century, and we can in principle relate any changes in class inequalities to such major educational reforms as the 1907 Free Place Regulations, the 1944 Education Act and the 1965 circular which initiated the replacement of a selective system of schooling with comprehensive schools. Educational researchers had great hopes of the 1944 reforms. Thus David Glass, the senior author of the first national study, was in no doubt that the Act would 'greatly increase the amount of social mobility in Britain' (Glass (1954), p. 22). 'Given the diminishing importance of economic and social background as a determinant of

Time Series: Theory and Methods

Abstract: 3. Time Series: Theory and Methods. By P. J. Brockwell and R. A. Davis. ISBN 0 387 96406 1. Springer, New York, 1987. x + 520pp. DM 120.

Chance or Chaos

Abstract: In my survey of mathematical statistics in 1940 the concept of chance was assigned a central role. Some remarks are now made on the comparatively new concept of chaos (which can arise in non-linear systems even when 'deterministic'), first of all in relation to two papers in epidemiology by Schaffer (1985) and Olsen et al. (1988), and then more generally on the recognition and properties of chaos and its relation with chance.

Health in Childhood and Social Inequalities in Health in Young Adults

Abstract: The role of childhood and adolescent health and development has been assessed as a potential explanation for later social class inequalities in health using longitudinal data from the 1958 cohort study. In early adulthood, general health (as self‐rated by respondents) and emotional health (as measured by the malaise inventory) were generally poorer in lower social classes. Adjustment for a wide range of childhood health characteristics did not reduce class differences in health evident in early adulthood and were therefore considered to be relatively unimportant influences. Some reduction in the class differences did occur, however, when allowance was made for ill‐health in adolescence. For example, odds of ‘poor’ or ‘fair’ rating of health at age 23 years in classes IV and V relative to classes I and II decreased from 2.52 to 2.26 in men and 3.39 to 3.08 in women after adjusting for school absence through ill‐health at age 16. Adolescent health may contribute to class inequalities in health but preliminary consideration of other potential influences (socioeconomic circumstances and health‐related behaviour) suggests that other factors may be relatively more important.

Unified Theory and Strategies of Survey Sampling

Abstract: Part A: Unified Theory of Sampling. Introduction. Principal Notation and Formulation of the Main Problem. Design-Based Estimation. Admissibility and Other Optimality Properties of Sampling Designs. Sufficiency and Related Concepts in Survey Sampling. Sample Survey and General Statistical Model. Further Details on Likelihood and Bayesian Approach. Predictive Approach. Robustness. Concluding Remarks. Appendices: An Optimality Property of the Sample Mean. Alternative Models and Relative Performances of Sampling Strategies. Inference About Symmetric Functions of Exchangeable Populations: Non-Informativeness of Labels. Robustness of Strategies. Part B: Strategies of Sampling. Introduction and Summary. The Horvitz-Thompson Estimator. Subdivision of Sampling Schemes. Sampling Schemes for Use of the HTE: n=2. Sampling Schemes for Use of the HTE: General n. Sampling with Other Strategies. Alternative Estimators. Ratio, Product, and Regression Estimators. Variance Estimation. Special Designs. Comparison of Strategies. Appendices: Varying Probability Sampling for Robust Estimation. Optimality of HTE Under Random Permutation Labelling. References.

Parental Education, Social Class and Entry to Higher Education 1976–86

Abstract: SUMMARY This paper uses the Scottish School Leavers' Survey series to estimate the net effects of parental education and parental social class, first, on a child's attainment of qualifications for entry to higher education and, second, on the chances of entry to higher education conditional on qualifying. Attention is given both to the university sector and to degree and advanced non-degree courses in the public sector. The findings confirm those of earlier studies but show in particular that the contribution of parental education to higher education entry is robust and has a basis in social and educational change that is wider than previously suspected. Projections of student demand that ignore these features are likely to be underestimates.

Accelerated Life Testing and Experts' Opinion in Reliability

Abstract: Introduction. Accelerated life testing and experts' opinion in reliability (D.V. Lindley). Experts' Opinions. Modern aspects of probability and utility (M.H. De Groot). The use of probability statements (D.V. Lindley). Choosing among experts (M.H. De Groot). A Bayesian view of weighted distributions and selection models (M.J. Bayarri and M.H. De Groot). Symmetry conditions on opinion assessment leading to time-transformed exponential models (F. Spizzichino). On the concept of dependence in probabilistic safety analysis and reliability (G. Apostolakis). The use of expert opinion in reliability: a survey (N.D. Singpurwalla and R. Soyer). Expert judgement in probabilistic safety assessment (G. Apostolakis). Expert opinions in software reliability analysis (S. Bologna and C.A. Clarotti). Using influence diagrams (R.E. Barlow). An application of influence diagrams (R.E. Barlow, R.W. Mensing, N.G. Smiriga). Accelerated Life Testing. Non-Bayesian approach to accelerated life testing: a short overview (C.A. Clarotti). Inferences from accelerated life tests. Some recent results (T.A. Mazzuchi and N.D. Singpurwalla). Bayesian design and analysis of accelerated life testing with step stress (M.H. De Groot and P.K. Goel). A Bayesian analysis of the stress-rupture life of Kevlar/epoxy spherical pressure vessels (R.E. Barlow, R.H. Toland and T. Freeman).

The Early Careers of Social Science Graduates and the Value of a Phd

Abstract: To look at the relevance of a PhD to the careers of graduates in the social sciences, the Economic and Social Research Council commissioned a survey of a sample of graduates in certain social science subjects who gained firstor upper second-class honours degrees at British universities between 1972 and 1977. A substantial proportion of respondents with PhDs, and especially of those who also held first-class honours, had become university teachers. The rest were scattered over a wide range of jobs, but few of them had needed a PhD to gain their first job after completing their studies, and there was no job group where the majority of respondents said that they had needed a PhD to do their first job well. In all job groups there were some respondents who regarded skills in research as relevant to their jobs, but other qualities imparted by a university, such as a training of the mind, were generally regarded as more relevant. Graduates with no university level post-graduate qualifications were generally paid more than those with a PhD, implying that employers put a higher value on experience in employment than on post-graduate research.

Lenin as a Statistician: a Non-Soviet View

Abstract: SUMMARY Many of Lenin's arguments were based on statistical analysis, and in the Soviet milieu he is widely praised as a statistician. In this paper we first examine the background and influences of Lenin's statistics and also consider his statistical arguments, some of which are often cited in the Soviet literature. Our general conclusions are that (a) in his earlier work Lenin exhibited innate craftmanship in analysing large-scale statistical data, (b) his statistics owes much to V. E. Postnikov, P. N. Skvortsov and F. A. Shcherbina whose influence has been downplayed, (c) Lenin was not beyond turning his analysis to political ends, (d) his influence among established statisticians in the Russian Empire, including the zemstvo statisticians, was quite modest and (e) he was instrumental in establishing official statistical organs in the USSR.