Moo-cow! Mummy! More! How do children learn so many words?

Learning the words of your language is a huge challenge for children. In particular, it is a challenge of scale. At least 40,000 words have to be learnt. At 18 months a child's vocabulary starts to explode. How do they do it? Bob McMurray says it is all to down to statistics.     

A model of regression lines through a common point: estimation of the focal point in wind-blown sand phenomena

This paper discusses a model in which the regression lines will be passing through a common point. This point exists as a focal point in the wind-blown sand phenomena. The model of regression lines will be called ‘the focal point regression model’. The focal point will move according to the conditions of the experiments or the measurement site, so it must be estimated together with regression coefficients. The existence of the focal point is mathematically proved in the research field of coastal engineering, but its physical meaning and exact estimation method have not been established. Considering the experimental and/or measurement conditions, five models, that is, common or different error variance(s), passing through or not the centroid and Bayes-like approach are proposed. Moreover, the formulae of direct computation for a focal point under some conditions are given for engineering purpose. The models are applied to the wind-blown sand data, and behaviors of the models are verified by numerical experiments.     

Survival in the U.S. petroleum refining industry

Of the 324 petroleum refineries operating in the U.S. in 1982, only 149 were still in the hands of their original owners in 2007. Using duration analysis, this paper explores why refineries change ownership or shut down. Plants are more likely to ‘survive’ with their original owners if they are older or larger, but less likely if the owner is a major integrated firm, or the refinery is a more technologically complex one. This latter result differs from existing research on the issue. This paper also presents a split population model to relax the general assumption of the duration model that all refiners will eventually close down; the empirical results show that the split population model converges on a standard hazard model; the log-logistic version fits best. Finally, a multinomial logit model is estimated to analyze the factors that influence the refinery plant's choices of staying open, closing, or changing ownership. Plant size, age and technology usage have positive impacts on the likelihood that a refinery will stay open, or change ownership (rather than close down).     

Challenges and Opportunities for Statistics and Statistical Education: Looking Back,Looking Forward

The 175th anniversary of the ASA provides an opportunity to look back into the past and peer into the future. What led our forebears to found the association? What commonalities do we still see? What insights might we glean from their experiences and observations? I will use the anniversary as a chance to reflect on where we are now and where we are headed in terms of statistical education amidst the growth of data science. Statistics is the science of learning from data. By fostering more multivariable thinking, building data-related skills, and developing simulation-based problem solving, we can help to ensure that statisticians are fully engaged in data science and the analysis of the abundance of data now available to us.     

Seeing the woods — statistics for the very young

If you go down to the woods today … you're sure of a big statistical experience. You might run into Chris Du Feu giving 9-year-olds their first lesson in statistics. Tasks for the teacher? One, clear nettles; two, make sure that leaves will stay firmly stuck in place on the bar-chart…     

Panic now! Or later?

Simon Briscoe , author of a new book, Panicology , examines the role that data plays in media scare stories. Scary newspaper stories are scattered throughout this piece. Just how much ought we to be panicked by headlines?     

Don't mention Stonehenge!

It rises massive and magnificent above Salisbury Plain. The extraordinary monoliths of Stonehenge leave us as impressed today as ever—and as baffled. Archaeologists argue about how it was made. Can statisticians tell the answer? Chris Jones thinks they can be of considerable assistance.     

Real data, real learning and the London Olympics

The Olympic and Paralympic Games are coming to London in 2012, and there will be huge interest, especially among the young. Could the Games be used to involve students of all ages in a large scale project that involves and interests them to break down their fear of statistics and to motivate learning? Neville Davies of the Royal Statistical Society Centre for Statistical Education appeals for help in a major teaching and learning initiative based on the London Olympics. Significance is proud to announce the launch of a scheme that could bring thousands of students, nationally and internationally, to appreciate and value the usefulness of statistics.     

目光深远的统计学大师 ——张尧庭先生逝世五周年纪念

 张尧庭先生是一位目光深远的统计学大师。本文认为他生前提出的下述观点即“统计学的发展全靠数学与实际问题的推动”,“为反映整体的、动态的东西统计学必须要有新突破”,以及“促进数理统计和经济统计的学术交流”等,在今天仍然具有现实意义,可为我们现时的统计学教学与研究的深入发展提供有益参考。     

Variance of the asymmetric n-player gambler’s ruin problem with ties allowed

In this article, the variance of the duration of play in the asymmetric n-player gambler’s ruin problem is considered, when the players use equal initial fortunes of d dollars, 1 ? d ? n + 1, and ties allowed in each round. Some special games are simulated and the simulation results verify the validity of the proposed formulas. It is shown that when we do not have the possibility of tie in the game, the increase in the number of players will change the ruin time from a random variable to a degenerate random variable. Finally, the three-tower problem with one of its different definitions are introduced and their expected times as well as their variances of the duration are considered.     

LSE method using the CHSS model

Testing the reliability at a nominal stress level may lead to extensive test time. Estimations of reliability parameters can be obtained faster thanks to step-stress accelerated life tests (ALT). Usually, a transfer functional defined among a given class of parametric functions is required, but Bagdonavi?ius and Nikulin showed that ALT tests are still possible without any assumption about this functional. When shape and scale parameters of the lifetime distribution change with the stress level, they suggested an ALT method using a model called CHanging Shape and Scale (CHSS). They estimated the lifetime parameters at the nominal stress with maximum likelihood estimation (MLE). However, this method usually requires an initialization of lifetime parameters, which may be difficult when no similar product has been tested before. This paper aims to face this issue by using an iterating least square estimation (LSE) method. It will enable one to initialize the optimization required to carry out the MLE and it will give estimations that can sometimes be better than those given by MLE.     

Just say no to data listings!

Sponsor companies often create voluminous static listings for Clinical Study Reports (CSRs) and regulatory submissions, and possibly for internal use to review participant-level data. This is likely due to the perception that they are required and/or lack of knowledge of various alternatives. However, there are other ways of viewing clinical study data that can provide an improved user experience, and are made possible by standard data structures such as the Study Data Tabulation Model (SDTM). The purpose of this paper is to explore some alternatives to providing a complete set of static listings and make a case for sponsors to begin considering these alternatives. We will discuss the recommendations from the PHUSE white paper, “Data Listings in Clinical Study Reports.”     

My statistician could have painted that! A statistical inquiry into modern art

How good are modern artists? Is great abstract art by modern masters a true expression of genius, or could a child, or a chimpanzee—or a statistician—do better? Mikhail Simkin makes a statistical inquiry into masterpieces and doodles to find out.     

Bootstrap-based ARMA order selection

Modelling the underlying stochastic process is one of the main goals in the study of many dynamic phenomena, such as signal processing, system identification and time series. The issue is often addressed within the framework of ARMA (Autoregressive Moving Average) paradigm, so that the related task of identification of the ‘true’ order is crucial. As it is well known, the effectiveness of such an approach may be seriously compromised by misspecification errors since they may affect model capabilities in capturing dynamic structures of the process. As a result, inference and empirical outcomes may be heavily misleading. Despite the big number of available approaches aimed at determining the order of an ARMA model, the issue is still open. In this paper, we bring the problem in the framework of bootstrap theory in conjunction with the information-based criterion of Akaike (AIC), and a new method for ARMA model selection will be presented. A theoretical justification for the proposed approach as well as an evaluation of its small sample performances, via simulation study, are given.     

Chaos: But Not in Both Directions!

This paper will informally explore the reversal of some stochastic autoregressive processes, which lead to deterministically chaotic processes. Correspondingly, the stochastic reversal of map models is shown to lead to a new class of invariant distribution. Finally, some connections between congruential recursions and independence in discretized chaotic processes are illustrated.     

中国股市：规律与风险——统计思维中的股市运行

中国股市：规律与风险——统计思维中的股市运行毛迅ＡＢＳＴＲＡＣＴＡｒｅｔｈｅｒｅａｎｙｒｅｇｕｌａｒｉｔｙｉｎｓｅｃｕｒｉｔｉｅｓ？ａｎｄｈｏｗｔｏｈｏｌｄｉｔ？Ｔｈｅａｕｔｈｏｒｃｏｎｓｉｄ－ｅｒｓｔｈａｔａｌｔｈｅｅｃｏｎｏｍｉｃｐｈｅｎｏｍｅｎ...     

First German register based census in 2011

Summary The last censuses in Germany were accomplished in 1987 (former federal states) and 1981 (new federal states), respectively. Germany was the only country which did not participate in the EU-wide census in 2000. A traditional census was not feasible on political scale due to high costs and low acceptance by the population. Complete, precise and actual statistical registers as a condition for a register based census, are non-existent in Germany. Official statisticians developed a model for a register based census with interviewer-based surveys kept to a minimum. As the basic conditions were not favourable the model became very complex and many methods and procedures had to be developed for the special case of German registers. Core model elements were tested in a census test, such as quality of registers (e. g. municipal population registers, registers of the federal labour agency) and practicability of procedures (e. g. merging of registers, household generation). The test result was that a register based census is possible but much development work still has to be done, as the registers do not have an acceptable quality yet. Additional interviewer based surveys will be necessary for correction of errors of the register and survey of additional variables. This is also true if measures to improve municipal population registers, e. g. a feedback procedure and a nationwide tax number are taken into account. So these measures will not substitute the additional interviewer based surveys. Provided that the essential investments in development and preparation will be performed, Germany could participate in the 2011 EU-census-round with a register based census at a fraction of the cost of a traditional census. Finally, the census 2011 will show the real quality of the registers, how the model will work throughout the country and which measures have to be taken before the following census.

---

Zusammenfassung In Deutschland wurde die letzte Volksz?hlung 1987 (Alte Bundesl?nder) bzw. 1981 (Neue Bundesl?nder) durchgeführt. An dem EU-weiten Zensus um das Jahr 2000 beteiligte sich Deutschland als einziger EU-Staat nicht, da ein herk?mmlicher Zensus aus Kosten- und Akzeptanzgründen politisch nicht durchsetzbar war. Die Voraussetzungen für einen Registerzensus, d. h. Statistikregister, die alle ben?tigten Merkmale mit hinreichender Genauigkeit und Aktualit?t enthalten, waren und sind nicht vorhanden. Amtliche Statistiker entwickelten ein Modell für einen registergestützten Zensus mit einem m?glichst geringen Anteil an prim?rstatistischen Erhebungen. Aufgrund der für einen Zensus auf Basis von Registern ungünstigen Rahmenbedingungen wurde dieses Modell relativ komplex und in vielen Bereichen statistisches Neuland betreten. Wesentliche Elemente des Modells wurden in einem sog. Zensustest überprüft. Getestet wurden insbesondere die Qualit?t und Aktualit?t vorhandener Register, wie Melderegister und Dateien der Bundesagentur für Arbeit, sowie die Eignung und Wirksamkeit der Verfahren, wie Zusammenführungen der Daten aus verschiedenen Quellen und Haushaltegenerierung. Die Tests best?tigten die Machbarkeit des Zensusmodells, zeigten aber gleichzeitig, dass noch umfangreiche Entwicklungsarbeiten erforderlich sind und dass die Register noch keine Qualit?t aufweisen, die einen Verzicht auf prim?rstatistische Erhebungen zur Korrektur und Erg?nzung der Registerergebnisse zulassen. Dies gilt auch unter Berücksichtigung der in den n?chsten Jahren vorgesehenen Ma?nahmen zur Verbesserung der Qualit?t der Melderegister, wie die Einführung eines elektronischen Rückmeldeverfahrens oder die Erg?nzung der Melderegister um eine bundeseinheitliche Steuernummer für jede Person. Soweit in den n?chsten Jahren die erforderlichen Investitionen in Entwicklung und Vorbereitung eines registergestützten Zensus get?tigt werden, k?nnte in Deutschland im Rahmen der n?chsten EU-Zensusrunde ein solcher Zensus im Jahre 2011 zu einem Bruchteil der Kosten eines herk?mmlichen Zensus durchgeführt werden. Erst mit diesem Zensus k?nnte dann letztlich festgestellt werden, welche Qualit?t die Register zu diesem Zeitpunkt aufweisen, wie sich das Modell eines registergestützten Zensus (fl?chendeckend) in der Praxis bew?hrt und welche Ma?nahmen für darauf folgende Zensen in Angriff genommen werden müssen.

     

AN EXTENSION OF THE EM ALGORITHM FOR OPTIMIZATION OF CONSTRAINED LIKELIHOOD: AN APPLICATION IN TOXICOLOGY

Since its introduction in the mid 1970's, the EM algorithm has found a widespread popularity for solving the likelihood equations. Several investigators have used the algorithm in a variety of problems with incomplete information to obtain maximum likelihood estimates in numerous applications. The algorithm, however, becomes inappropriate when the underlying equations are subject to some constraints. Although an extension has been proposed to derive solutions when the parameters are subject to a set of linear constraints, the evaluation of likelihood equations from incomplete data when the equations are subject to a nonlinear constraint is still an open problem. Here, we consider a mixture model, a classical example of incomplete data, and discuss the problem of maximum likelihood estimation of the model parameters when the parameters have to satisfy a nonlinear constraint. An extension of the EM algorithm based on the celebrated Lagrange multiplier will be proposed to solve the equations. An application of the methodology in animal bioassay experiments for risk assessment of toxic substances will be described and data from a toxicological experiment will be used to illustrate the results.     

ANASE: measuring aircraft noise annoyance very unreliably

Does anyone who lives under a flight-path like aircraft noise? It is a political hot potato as well as a peace-destroyer. Tens of thousands of people will hear the noise from any third runway at Heathrow. So, when a study commissioned by the government claimed that people are becoming less tolerant of aircraft noise, it made highly unpleasant reading for supporters of a third runway. But the Department for Transport rejected the report as unreliable. Peter Brooker senses the vibrations.     

On robustness of optimal balanced resolution v plans

In this paper, we present Srivastava-Chopra optimal balanced resolution V plans for 2^m factorials (4?m?8) which are robust in the sense that, when any observation is missing, each of these designs will remain as a resolution V plan.