David Cox

• Nuffield College, Oxford

A broad account is given of statistical significance tests outlining their role and their limitations in the careful analysis of data.

Objective: Cystic Fibrosis related diabetes (CFRD) is a risk factor for mortality, with pathogenesis possibly related to susceptibility of CFTR deficient pancreatic b-cells to oxidative stress. Because airway infections may increase oxidative stress in CF patients, we evaluated their association with CFRD. Methods: We analyzed 23,622 patients with...

A broad review is given of the general principles underlying study design with emphasis on applications in medical and epidemiological contexts. The main theme of the paper is that, while the distinction between interventionist studies, that is experiments, and purely observational ones is important, there are many common threads. A wide range of s...

A theoretical analysis is made of the properties of various methods for comparing two distributions of survival time. The results are intended primarily to guide the choice of method of analysis for such simple comparisons as of a treatment versus a control, but the main implications are fairly general, illustrating the performance of different mod...

Principles for the planning and analysis of observational studies, as suggested by W.G Cochran in 1972, are discussed and compared to additional methodological developments since then.

Editors' Note: This conversation covers the extensive contributions of John C. Gower to statistical methodology, computing and applications. It also touches on the pioneering development of statistical computing at Rothamsted Experimental Station and the relevant work of Frank Yates, John Nelder and others who were at Rothamsted. See also the accom...

Statistical theory aims to provide a foundation for studying the collection and interpretation of data, a foundation that does not depend on the particular details of the substantive field in which the data are being considered. This gives a systematic way to approach new problems, and a common language for summarising results; ideally, the foundat...

We describe how graphical Markov models started to emerge in the last 40 years, based on three essential concepts that had been developed independently more than a century ago. Sequences of joint or single regressions and their regression graphs are singled out as being best suited for analyzing longitudinal data and for tracing developmental pathw...

Objective: Show relative survivorship associated with specific disease-causing mutations of the cystic fibrosis (CF) transmembrane conductance regulator (CFTR). Methods: We assessed survival effects of CFTR mutations using the CF Foundation Patient Registry (CFFPR). We estimated hazard ratios corrected for age and sex for F508del compound heterozyg...

Two papers of 50 years ago concerning the comparison of non-nested models are reviewed and some subsequent developments and problems are sketched.

Background: Spontaneous coronary artery dissection (SCAD) is an increasingly recognized cause of acute coronary syndrome, predominately in women of child bearing age. They are identified primarily by clinical markers of acute coronary syndrome (ACS) and angiographic evidence suggesting external lumen compression from medial hema-toma. Optimal manag...

Badgers are involved in the transmission to cattle of bovine tuberculosis (TB), a serious problem for the UK farming industry. Cross-sectional studies have shown an association between bite wounds and TB infection in badgers which may have implications for M. bovis transmission and control, although the sequence of these two events is unclear. Tran...

Consider parametric models that are too complicated to allow calculation of a likelihood but from which observations can be simulated. We examine parameter estimators that are linear functions of a possibly large set of candidate features. A combination of simulations based on a fractional design and sets of discriminant analyses is then used to fi...

Background: Patients with diabetes mellitus (DM) undergoing stent implantation have increased on-treatment platelet reactivity (HPR), which may contribute to a greater risk of post-procedural thrombotic events.

One attractive feature of optimum design criteria, such as D-and A-optimality, is that they are directly related to statistically interpretable properties of the designs that are obtained, such as minimizing the volume of a joint confidence region for the parameters. However, the assumed relationships with inferential procedures are valid only if t...

We present the results of a 2005 case-control study of bovine tuberculosis (bTB) breakdowns in English and Welsh herds. The herd management, farming practices, and environmental factors of 401 matched pairs of case and control herds were investigated to provide a picture of herd-level risk factors in areas of varying bTB incidence. A global conditi...

Whether thrombus aspiration and local glycoprotein IIb/IIIa administration reduce infarct size in patients with ST-elevation myocardial infarction (STEMI) undergoing primary percutaneous coronary intervention (PCI) has not been established in multicenter studies. INFUSE-AMI is a multicenter, open-label, controlled, single-blind randomized study enr...

Our objective was to evaluate the impact of door-to-balloon time (DBT) on mortality depending on clinical risk and time to presentation. DBT affects the mortality rate in ST-segment elevation myocardial infarction treated with primary percutaneous coronary intervention, but the impact may vary across subgroups. The CADILLAC (Controlled Abciximab an...

Lung transplantation is a complex, high-risk, potentially life-saving therapy for the end-stage lung disease of cystic fibrosis (CF). The decision to pursue transplantation involves comparing the likelihood of survival with and without transplantation as well as assessing the effect of wait-listing and transplantation on the patient's quality of li...

A general review is given of the role of randomization in experimental design. Three objectives are distinguished, the avoidance of bias, the establishment of a secure base for the estimation of error in traditional designs, and the provision of formally exact tests of significance and confidence limits. The approximate randomization theory associa...

This fine paper 1 is surely of direct value to all interested in the history of a major issue, possibly the major issue, in non-infectious disease epidemiology. It also sends a strong message to epidemiologists, statisticians and those from the machine learning world who are concerned with potential causal interpretation of their data. It may indee...

In most social animals, the prevalence of directly transmitted pathogens increases in larger groups and at higher population densities. Such patterns are predicted by models of Mycobacterium bovis infection in European badgers ( Meles meles ). We investigated the relationship between badger abundance and M. bovis prevalence, using data on 2696 adul...

Square contingency tables with matching ordinal rows and columns arise in particular as empirical transition matrices and the paper considers these in the context of social class and income mobility tables. Such tables relate the socio-economic position of parents to the socio-economic position of their child in adulthood. The level of association...

Bovine tuberculosis (TB) is a zoonotic disease that can have serious consequences for cattle farming and, potentially, for public health. In Britain, failure to control bovine TB has been linked to persistent infection of European badger (Meles meles) populations. However, culling of badgers in the vicinity of recent TB outbreaks in cattle has fail...

We introduce and study distributions of sets of binary vari-ables that are symmetric, that is each has equally probable levels. The joint distribution of these special types of binary variables, if generated by a recursive process of linear main effects is essentially parametrized in terms of marginal correlations. This contrasts with the log-linea...

If two parameters ψ and λ, are orthogonal, λψ, the maximum-likelihood estimate of λ for given ψ, varies only slowly with ψ in the neighbourhood of the overall maximum-likelihood point. The same is true if ψλ is replaced by a nonlinear function h(ψλ). The detailed form of the variation of ψλ with ψ is studied, and a basis suggested for choosing a pa...

It is common, in particular in observational studies in epidemiology, to impose stratification to adjust for possible effects of age and other variables on the binary outcome of interest. Overstratification may lower the precision of the estimated effects of interest. Understratification risks bias. These issues are studied analytically. Asymptotic...

Undetected confounding may severely distort the effect of an explanatory variable on a response variable, as defined by a stepwise data-generating process. The best known type of distortion, which we call direct confounding, arises from an unobserved explanatory variable common to a response and its main explanatory variable of interest. It is rele...

We provide brief comments on some common threads arising from the 'wishlists' set out in some of the other papers in this volume. The discussion is necessarily incomplete: in particular we have dealt only with points for which a reasonably compact answer seems possible.

The link between tuberculosis in cattle and in badgers has long been contentious. The Independent Scientific Group (ISG) on Cattle TB, a group containing some of the country's top statisticians, investigated it. In June this year, after 9 years of work, it reported to Ministers that a cull of badgers would be unlikely to contribute usefully to the...

Several different types of statistical interaction are defined and distinguished, primarily on the basis of the nature of the factors defining the interaction. Illustrative examples, mostly epidemiological, are given. The emphasis is primarily on interpretation rather than on methods for detecting interactions. Comment: Published in at http://dx.do...

The Randomized Badger Culling Trial (RBCT) began in 1998 to determine the impact of badger culling in controlling bovine tuberculosis in cattle. A total of 1166 badgers (14% of total) proactively culled during the RBCT were found to be tuberculous, offering a unique opportunity to study the pathology caused by Mycobacterium bovis in a large sample...

Letter focusing on the randomised badger culling trial (RBCT) made by Gallagher and others (2007). While the authors contrast the RBCT findings with the results of ‘previous trials in this country’, in fact no scientifically designed controlled trials of badger culling had been conducted in Britain before the RBCT. The letter states the RBCT findin...

To determine if outcomes could be further improved, we investigated an embolic protection device placed proximal to the target lesion that could provide protection before lesion instrumentation, allow the use of conventional guidewires, and permit embolic protection in anatomy unfavorable for distal devices. Embolic complications during stenting of...

Bovine tuberculosis (TB), caused by Mycobacterium bovis, has serious consequences for Britain's cattle industry. European badgers (Meles meles) can transmit infection to cattle, and for many years the British government culled badgers in a series of attempts to reduce cattle infections. We investigated the impact of badger culling on the spatial...

A brief review is given of procedures for the collective analysis of a large number of significance tests. A simple procedure previously supplied for isolating 'real' effects on the basis of a large number of significance tests is generalized to deal with two-sided tests and is also related more explicitly to the false discovery rate.

The main phases of applied statistical work are discussed in general terms. The account starts with the clarification of objectives and proceeds through study design, measurement and analysis to interpretation. An attempt is made to extract some general notions. Comment: Published at http://dx.doi.org/10.1214/07-AOAS113 in the Annals of Applied Sta...

A relationship due to W. G. Cochran [J. R. Stat. Soc., Suppl. 5, 171–176 (1938; Zbl 0019.31902)] showing the effect on least squares regression coefficients of marginalizing over or conditioning on an explanatory variable is generalized to quantile regression coefficients. The condition under which conditioning does not induce interaction or effect...

Culling is often considered as a tool for controlling wildlife diseases that can also infect people or livestock. Culling European badgers Meles meles can cause both positive and negative effects on the incidence of bovine tuberculosis (TB) in cattle. One factor likely to influence the outcome of different badger-culling strategies for cattle TB is...

Bovine tuberculosis (TB) has re-emerged as a major problem for British cattle farmers. Failure to control the infection has been linked to transmission from European badgers; badger culling has therefore formed a component of British TB control policy since 1973. To investigate the impact of repeated widespread badger culling on cattle TB, the Rand...

Primer names and nucleotidic sequence used for ACCN1 exon resequencing. (0.05 MB DOC)

Minor allele frequency of studied SNPs (information from the founders of the Nuoro population and from the Caucasian population). (0.05 MB DOC)

Characteristics of the analyzed microsatellites. (0.06 MB DOC)

Primer names, PCR conditions and sequence conditions for the exons of the ACCN1 gene. (0.05 MB DOC)

D′ coefficient for each pair of studied SNPs (information from the founders of the Nuoro population and from the Caucasian population). (0.05 MB DOC)

R2 coefficient for each pair of studied SNPs (information from the founders of the Nuoro population and from the Caucasian population). (0.05 MB DOC)