Weighted rank statistics in factorial designs with fixed effects

Nonparametric methods for the analysis of factorial designs with two fixed factors are considered. The sample sizes may be unequal and the distribution functions are not assumed to be continuous. Nonparametric hypotheses for the main effects and for the interaction are tested by ranking procedures where the statistics are weighted according to the different sample sizes within the levels of one factor. Simulations show that the approximations by the limiting normal distribution and by the t - and F -distributions are quite accurate if the samples sizes within the cells are at least 7 for all treatment combinations. Moreover, it turns out that the power for the weighted statistics is much higher than for the unweighted statistics. The application of the suggested procedures is demonstrated by the analysis of a data set from a clinical trial with ordered categorical data.     

An Interview with Chris Skinner

Chris Skinner was born in London on 12 March 1953. He completed a BA in mathematics in 1975 at the University of Cambridge. He then obtained an MSc degree in statistics from the London School of Economics and Political Science (LSE) in 1976 and worked as an assistant statistician in the Central Statistical Office for 1 year. After working as a research assistant in LSE from 1977 to 1978, he joined the University of Southampton as a lecturer in 1978, where he earned a PhD in social statistics in 1982. He remained at the University of Southampton, where he became a senior lecturer in 1989 and professor of statistics in 1994. While serving as the head of his department from 1997 to 2000, he played a crucial role in the creation of an MSc programme in official statistics in 1999. In 2011, he returned to the LSE, where he currently holds the position of professor of statistics. Chris is the author of over 80 peer‐reviewed articles in statistical journals and the co‐editor of two influential books on the analysis of survey data. He made significant research contributions covering areas that include the analysis of survey data, inference in the presence of non‐response and measurement errors and statistical disclosure control. He served on several advisory committees, including the Statistical Methods Advisory Committee at Statistics Canada (from 2000 to 2011) and the National Statistics Methodology Advisory Committee in the United Kingdom (from 2001 to 2010). He has received numerous awards and honors for his outstanding contributions to survey sampling and social statistics. He is a Fellow of the American Statistical Association, Fellow of the British Academy and a Fellow of the Academy of Social Sciences. In 2009, he received the West Medal from the Royal Statistical Society for contributions to social statistics, and in 2010, he was made a Commander of the Most Excellent Order of the British Empire. In 2019, he also received the Waksberg award to recognize his contributions to survey methodology. The following conversation took place at LSE on 21 May 2019.     

Roles of Statistics in the Life Sciences

Quantitative methodologies have always been central in the life sciences, and some of these belong to the general area of statistics. Other methods of data analysis have their part to play, and it is of ongoing importance for the field of statistics to adapt to these developments.
Besides the role of statistical methods in generating new knowledge, there are also regulatory uses of statistics, particularly in drug registration and in environmental standard setting. These dual roles of statistical methods should not be confused.
Statistical data from official sources and elsewhere play important roles in the life sciences, and often necessitate very careful adherence to confidentiality.
The historical role of the ISI towards life sciences has been modulated by the existence of other international organizations in the area.     

Biased statistics with zero observations in multiple regression analysis of regional data

Observations containing zeroes as the values of all variables, and which are not meaningful, are shown to be likely in certain regional data sets which may be subjected to multiple regression analysis. The biasing effects of such observations on regression statistics are shown and illustrated with a small data set. It is recommended that such zero observation be identified and removed from regional data sets prior to analysis.     

Equity-premium prediction: Attention is all you need

Predictions of stock returns are greatly improved relative to low-dimensional forecasting regressions when the forecasts are based on the estimated factor of large data sets, also known as the diffusion index (DI) model. However, when applied to text data, DI models do not perform well. This paper shows that by simply using text data in a DI model does not improve equity-premium forecasts over the naive historical-average model, but substantial gains are obtained when one selects the most predictive words before computing the factors and allows the dictionary to be updated over time.     

Past absence as a predictor of present absence: the case of a large Danish municipality

This article investigates whether past absence behaviour is a predictor of present absence duration in a large Danish municipality with 17,499 individuals observed from 1996 to 2004. Past absence behaviour is measured in both absence days and absence spells. The article also investigates a number of confounders such as gender, age, seniority, wage, contracted number of work hours and season. The results of the empirical study show that there is a significant positive relationship between employees' absence duration and past absence spells and past absence days, respectively. The study thus confirms that past days and past spells have an equal potential of predicting present absent. Past absence behaviour can thus be used as an early warning for managers. The study also confirms that personal characteristics such as age and seniority also influence absence duration. Moreover, job characteristics such as wage and contracted number of work hours also influence absence duration. Finally, the season of the year seems to influence absence duration.     

On the use of grouping methods in the analysis of residential housing markets

In this article we employ four different statistical techniques (geographic, AID, cluster and discriminant analysis) to define homogeneous groupings of houses within an urban area. Analysis of a sample of data from Fayette Country, Kentucky indicates that each of these methods produces distinguishable homogeneous groupings of properties. Predictions of house values are compared using data from Lane County, Oregon, San Mateo County, California, and Fayette County. The major conclusions of the study are that there are no discernible differences among the four methods and that predictions made ignoring the grouping information are as accurate as those obtained by grouping.     

Network effects in influenza spread: The impact of mobility and socio-economic factors

This paper introduces new methods of modeling and analyzing social networks that emerge in the context of disease spread. Four methods of constructing informative networks are presented, two of which use. static data and two use temporal data, namely individual citizen mobility observations taken over an extensive period of time. We show how the built networks can be analyzed, and how the numerical results can be interpreted, using network permutation-based surprise analysis. In doing so, we explain the relationship of surprise analysis with conventional network hypothesis testing and Quadratic Assignment Procedure regression. Surprise analysis is more comprehensive, and can be without limitation performed with any form(s) of network subgraphs, including those with multiple nodal attributes, weighted links, and temporal features. To illustrate our methodological work in application, we put them to use for interpreting networks constructed from the data collected over one year in an observational study in Buffalo and Erie counties in New York state during the 2016–2017 influenza season. Even with the limitations in the data size, our methods are able to reveal the global (city- and season-wide) patterns in the spread of influenza, taking into account population mobility and socio-economic factors.     

Multi-source Statistics: Basic Situations and Methods

Many National Statistical Institutes (NSIs), especially in Europe, are moving from single-source statistics to multi-source statistics. By combining data sources, NSIs can produce more detailed and more timely statistics and respond more quickly to events in society. By combining survey data with already available administrative data and Big Data, NSIs can save data collection and processing costs and reduce the burden on respondents. However, multi-source statistics come with new problems that need to be overcome before the resulting output quality is sufficiently high and before those statistics can be produced efficiently. What complicates the production of multi-source statistics is that they come in many different varieties as data sets can be combined in many different ways. Given the rapidly increasing importance of producing multi-source statistics in Official Statistics, there has been considerable research activity in this area over the last few years, and some frameworks have been developed for multi-source statistics. Useful as these frameworks are, they generally do not give guidelines to which method could be applied in a certain situation arising in practice. In this paper, we aim to fill that gap, structure the world of multi-source statistics and its problems and provide some guidance to suitable methods for these problems.     

Software for Learning and for Doing Statistics

The community of statisticians and statistics educators should take responsibility for the evaluation and improvement of software quality from the perspective of education. The paper will develop a perspective, an ideal system of requirements to critically evaluate existing software and to produce future software more adequate both for learning and doing statistics in introductory courses. Different kinds of tools and microworlds are needed. After discussing general requirements for such programs, a prototypical ideal software system will be presented in detail. It will be illustrated how such a system could be used to construct learning environments and to support elementary data analysis with exploratory working style.     

Factors Related to the Academic Performance of Students in the Statistics Course in Psychology

Many studies have examined the factors that influence academic performance in primary and secondary education as well as at university, with the purpose of enhancing learning at these stages and reducing drop-out rates. It is within this research framework that we want to emphasise the deficient performance of students enrolled on the statistics course in the Faculty of Psychology at the University of Barcelona. Consequently, this paper attempts to determine the factors that affect student performance in this subject by undertaking an analysis of a structural equation model and determining its stability over time. In order to accomplish our objective, we worked with two samples of students enrolled statistics classes. The first group comprised 211 students enrolled in the academic year 2000–2001, while the second comprised 287 students enrolled in the academic year 2001–2002. By administering a questionnaire, we obtained information concerning such variables as demographic data, previous academic record, information related to the subject and the degree of satisfaction with it, and the final mark obtained by the students in the subject. The parameters for each group of students were estimated separately and the goodness of fit of the proposed structural model was assessed. The data analysis showed a good fit with both data bases, but the set of estimated parameters differed in the two academic years under consideration.     

Foundational Value of Statistics Education for Management Curriculum

The purpose of this paper is to propose a unique and distinct value of statistics education for management. The 1986 inaugural conference on Making Statistics More Effective in Schools of Business (MSMESB) proposed valuable guidelines for reforming statistics education in schools of business. However, a survey conducted by McAlevey & Everett (2001) identified that their impact has been minimal, and argued that structural problems many business schools have are the potential cause. We argue these structural problems exist because the value of the body of statistical tools for management is ambiguous and has not been made explicit. The unique and distinct value of statistics for management can be identified as the body of tools necessary to meet the inherent needs of a manager charged with making predictive judgments facing data. The need arises because human information‐processing capacity is quite limited, as the findings of researchers in cognitive psychology testify. These findings also affirm that the basic statistical concepts needed for processing data cannot be learned from management experiences. The model of a manager faced with data, while considering the evidence of inherent limitations of human information‐processing capacity, establishes the foundational value of statistics training in the management curriculum. Statistics education in business schools will be made more effective when management educators recognize such value of the discipline, lend their support and reward the ownership commitment for continuous improvement and innovations of the business statistics curriculum.     

一月攻略——留住你的员工

“经过慎重的考虑,我决定辞职,我找到另外一个机会,我接受了另外一份工作,我们能谈谈吗？”这是王杰在新年第一天给公司老总发的一封邮件。     

Aspects of statistical consulting not taught by academia

Education in statistics is preparing for statistical analysis but not necessarily for statistical consulting. The objective of this paper is to explore the phases that precede and follow statistical analysis. Specifically these include: problem elicitation, data collection and, following statistical data analysis, formulation of findings, and presentation of findings, and recommendations. Some insights derived from a literature review and real-life case studies are provided. Areas for joint research by statisticians and cognitive scientists are outlined.     

Towards an Integrated Statistical System at Statistics Netherlands*

Changes in circumstances put pressure on Statistics Netherlands (SN) to redesign the way its statistics are produced. Key developments are: the changing needs of data‐users, growing competition, pressure to reduce the survey burden on enterprises, emerging new technologies and methodologies and, first and foremost, the need for more efficiency because of budget cuts. This paper describes how SN, and especially its business statistics, can adapt to these new circumstances. We envisage an optimum situation as one with a single standardised production line for all statistics and a central data repository at its core. This single production line is supported by generic and standardised tools, metadata and workflow management. However, it is clear that such an optimum situation cannot be realised in just a few years. It should be seen as the point on the horizon. Therefore, we also describe the first transformation steps from the product‐based stovepipe‐oriented statistical process of the past to a more integrated process of the future. A similar modernisation process exists in the area of social statistics. In the near future both systems of business and social statistics are expected to connect at pivotal points and eventually converge on one overall business architecture for SN. Discussions about such an overall business architecture for SN have already been started and the first core projects have been set up.     

Disclosure‐protected Inference Using Generalised Linear Models

Vast amounts of data that could be used in the development and evaluation of policy for the benefit of society are collected by statistical agencies. It is therefore no surprise that there is very strong demand from analysts, within business, government, universities and other organisations, to access such data. When allowing access to micro‐data, a statistical agency is obliged, often legally, to ensure that it is unlikely to result in the disclosure of information about a particular person or organisation. Managing the risk of disclosure is referred to as statistical disclosure control (SDC). This paper describes an approach to SDC for output from analysis using generalised linear models, including estimates of regression parameters and their variances, diagnostic statistics and plots. The Australian Bureau of Statistics has implemented the approach in a remote analysis system, which returns analysis output from remotely submitted queries. A framework for measuring disclosure risk associated with a remote server is proposed. The disclosure risk and utility of approach are measured in two real‐life case studies and in simulation.     

Transferability of disaggregate mode choice models

Predictions of aggregate transport mode split for inter-city trips are derived from a disaggregate model of travel demand. A series of tests are performed to assess the limitations of the prediction methodology for disaggregate models, and it is shown that disaggregate models are capable of predictions across diverse travel situations. The disaggregate model predictions are compared with predictions derived from aggregate models such as are currently used in urban transportation planning, and it is shown that disaggregate models based on much smaller data sets predict better than aggregate models while requiring no more information about the predicted population.     

Epistemology of data envelopment analysis and comparison with other fields of OR/MS for relevance to applications

Periodically, professionals in a given field must reflect and assess where the field has been, where it is heading, and what if anything, should be done to change that field's course. This article discusses statistical trends within the data envelopment analysis (DEA) literature. The number of articles published per year in refereed journals over the entire lifespan of the field, authorship and publishing outlets-of-choice statistics are used to indicate DEA's vitality, relevance, diffusion to other disciplines/professions and its worldwide acceptance. Lastly, based on published meta reviews, comparisons are made with other OR/MS sub-disciplines.     

Influence measures on profile analysis with elliptical data through Frèchet’s metric

In this paper, from Frèchet’s metric, diagnostic tools are constructed for the detection of influential observations in Profile Analysis with elliptically distributed random errors . This distributional hypothesis allows the application of the proposed diagnostics to a wide variety of random experiences, not only for data from a multivariate normal distribution but also from other symmetric distributions, commonly used in studies of several sciences. The diagnostics are based on Frèchet’s distance between the distributions of the basic statistics in Profile Analysis, in the postulated model and in the perturbed model obtained by deleting an observation from the sample data. This metric is highly useful since it enables the analysis of the influence on the point estimation and the estimation error. Applications on two data sets are provided.     

A Two‐Step Procedure to Measure Representativeness of Internet Data Sources

So far, statistics has mainly relied on information collected from censuses and sample surveys, which are used to produce statistics about selected characteristics of the population. However, because of cost cuts and increasing non‐response in sample surveys, statisticians have started to search for new sources of information, such as registers, Internet data sources (IDSs, i.e. web portals) or big data. Administrative sources are already used for purposes of official statistics, while the suitability of the latter two sources is currently being discussed in the literature. Unfortunately, only a few papers devoted to statistical theory point out methodological problems related to the use of IDSs, particularly in the context of survey methodology. The unknown generation mechanism and the complexity of such data are often neglected in view of their size. Hence, before IDSs can be used for statistical purposes, especially for official statistics, they need to be assessed in terms of such fundamental issues as representativeness, non‐sampling errors or bias. The paper attempts to fill the first gap by proposing a two‐step procedure to measure representativeness of IDSs. The procedure will be exemplified using data about the secondary real estate market in Poland.