Vague prior: A term used for the prior distribution in Bayesian methods in the situation when there is complete ignorance about the value of a parameter. [American Journal of Epidemiology, 2005, 162, 694–703.]

Validity: The extent to which a measuring instrument is measuring what was intended, or the degree to which the inference drawn from a study is warranted.

Validity checks: A part of data editing in which a check is made that only allowable values or codes are given for the answers to questions asked of subjects. A negative height, for example, would clearly not be an allowable value.

Variable: Some characteristic that differs from subject to subject or from time to time.

Variance: A measure of the spread or dispersion of a random variable around its mean. Generally assessed by the sum of squared deviations of a set of sample observations from their arithmetic mean divided by n―1, where n is the sample size. This provides an unbiased estimator of the population value. [Altman, D. G., 1991, Practical Statistics for Medical Research, Chapman and Hall/CRC, Boca Raton, FL.]

Variance components: A term generally used for the variances of random effects in statistical models, for example mixed-effects models. Particularly important in quantitative genetics where phenotypic variation is often partitioned into genetic variation, environmental variation, and the interaction of genetic and environmental variation. [Searle, S.R., Casella, G. and McCulloch, C. E., 1992, Variance Components, J. Wiley & Sons, New York.]

Data and safety monitoring board: A committe set up to monitor a clinical trial for patient safety and for evidence of benefit. Needed to ensure minimal acceptable risks to trial participants and to continually reassess the risks versus benefits of trial interventions. [Circulation, 1995, 91, 901–4].

Data archives: Generally large sets of data that can be accessed and utilized by researchers intending to perform secondary data analysis. Such archives preserve important data against disposal or deterioration. [British Medical Journal, 1994, 308, 1519–20.]

Database: A structured collection of data that is organized in such a way that it may be accessed easily by a wide variety of applications programs. Large clinical databases are becoming increasingly available to clinical and policy researchers and are generally used for two purposes: to facilitate healthcare delivery and for research. An example of such a database is that provided by the US Health Care Financing Administration, which contains information about all Medicare patients' hospitalizations, surgical procedures and surgery visits. See also administrative databases. [Elmasri, R. and Navathe, S. B., 1994, Fundamentals of Database Systems, 2nd edn, Addison-Wesley, Reading, MA.]

Database management system: A computer system organized for the systematic management of a large, structured collection of information that can be used for storage, modification and retrieval of data. [Controlled Clinical Trials, 1995, 16, 30S–65S.]

Data dredging: A term used to describe comparisons made withing a data set not specifically described before the start of the study. [Altman, D. G., 1991, Practical Statistics for Medical Research, CRC/Chapman & Hall, London.]

G2: Often used to denote deviance.

Galbraith plot: A graphical method for identifying outliers in a meta―analysis. The standardized effect size is plotted against precision (the reciprocal of the standard error). If the studies are homogeneous, then they should be distributed within +/−2 standard errors of the regression line through the origin. An example is given in Figure 42. [Ader, H. J. andMellenberg, G. J., 1999, ResearchMethodology, Sage, London.]

GAM: Abbreviation for geographical analysis machine and for generalized additive model.

Gambler's fallacy: The belief that if an event has not happened for a long time, then it is bound to occur soon. [Everitt, B. S., 1999, Chance Rules, Springer, New York.]

Game theory: The branch of mathematics that deals with the theory of contests between two or more players under specified sets of rules. The subject assumes a statistical aspect when part of the game proceeds under a chance scheme. [International Journal of Game Theory, 1979, 8, 175–92.]

Gamma distribution: A probability distribution of which the shape depends on two parameters. Some examples are given in Figure 43. Often used to model heavily skewed data. [Evans, M., Hastings, N. and Peacock, B., 2000, Statistical Distributions, 3rd edn, J.Wiley & Sons, New York.]

Garbage in, garbage out (GIGO): A term that draws attention to the fact that sensible output follows only from sensible input. Specifically, if the data are originally of dubious quality, then so will be the results.

Back-calculation: A method of estimating past infection rate of an epidemic infectious disease by working backwards from observed disease incidence rate using knowledge of the incubation period between infection and disease. Used mainly for reconstructing plausible HIV incidence curves from AIDS incidence data. Limitations of the approach are that it provides little information about recent infection rates and that projections can be sensitive to recent incidence. [Statistical Science, 1993, 8, 82–119.]

Background level: The usually low concentration of some substance or agent that is characteristic of a particular time or place rather than a specific hazard. An example is the background level of naturally occurring forms of ionizing radiation to which nearly everybody is exposed.

Back-projection: Synonym for back-calculation. Backward elimination: See selection methods in regression.

Backward-looking study: An alternative (and unattractive) term for retrospective study.

Balaam's design: A design for testing differences between two treatments, A and B, in which patients are allocated randomly to one of four sequences, AA, AB, BA or BB. See also crossover design. [Statistics in Medicine, 1988, 7, 471–82.]

Balanced design: A term usually applied to any experimental design in which the same number of observations is taken for each combination of the experimental factors. See also non-orthogonal design. Balanced incomplete block design: An experimental design in which not all treatments are used in all blocks. Such designs have the following properties:

• Each block contains the same number of units.

• Each treatment occurs the same number of times in all blocks.

• Each pair of treatment combinations occurs together in a block the same number of times as any other pair of treatments.

Tango's index: An index for summarizing the occurrences of cases of disease in a stable geographical unit where the occurrences are grouped into discrete intervals. Can be used to detect disease clusters occurring over time. See also ratchet scan statistic and scan statistic. [Statistics in Medicine, 1993, 12, 1813–28.]

Target population: The collection of individuals, items, measurements, etc. about which it is required to make inferences. Often, the population that is actually sampled differs from the target population, which may result in misleading conclusions being made. For example, if an investigator is interested in some aspect of characterizing the natural history of rheumatoid arthritis, then the target population will be all patients with the disease. If, however, the investigator examines only those cases arising from, say, the records office of a large university hospital, then it is likely that selective factors will cause the population that is actually sampled to differ from the target population for a number of reasons, including the fact that rheumatoid arthritis does not always require hospitalization. [Colton, T., 1974, Statistics in Medicine, Little, Brown and Company, Boston, MA.]

TD50: Abbreviation for tumorigenic dose 50.

Telephone sampling: The use of a telephone for sample survey data collection. Telephone surveys are commonly used in market research, primarily because good samples can be achieved at reasonable cost. Sampling can be via directories, which can lead to bias because many households are unlisted, or random digit dialling, in which random four digit numbers in known exchanges are generated to assure that both listed and unlisted households are included.

Zelen's single-consent design: An alternative to simple random allocation for forming treatment groups in a clinical trial. Begins with the set of N eligible patients. All N of these patients are then subdivided randomly into two groups, say groups G1 and G2of sizes n1 and n2. The standard therapy is applied to all the patients assigned toG1. The new therapy is assigned only to those patients in G2 who consent to its use. The remaining patients who refuse the new treatment are treated with the standard therapy. The main advantages of the design are that almost all eligible individuals are included in the trial and that it allows the evaluation of the true effect of offering experimental interventions to patients. The main disadvantages are that such trials have to be open-label trials and that the statistical power of the study may be affected as a high proportion of participants choose to have the standard treatment. [New England Journal of Medicine, 1979, 300, 12425]

Zero-sum game: A game played by a number of people in which the winner takes all the stakes provided by the losers so that the algebraic sum of gains at any stage is zero. Many decision problems can be modelled as such games involving two people. [New Scientist, 1990, 4, 1.]

Z-scores: Synonym for standard scores.

z-test: A test for assessing the hypothesis that the mean of a normal distribution takes a particular value, or for assessing whether the means of two normal distributes with the some variance are equal.

Mack―Wolfe test: A distribution-free method for one-way designs used to test a null hypothesis of equality of treatment effects against an alternative specifying an umbrella ordering. [Hollander, M. andWolfe, D. A., 1999, Nonparametric Statistical Methods, J. Wiley & Sons, New York.]

Mahalanobis D2: A measure of the distance between two populations or two samples of individuals based on observations on a number of variables measured on each. The measure is based on the difference of the mean vectors of each group and on their assumed common variance―covariance matrix. See also Hotelling's T test. [Everitt, B. S. and Dunn, G., 2001, Applied Multivariate Data Analysis, 2nd edn, Arnold, London.]

Main effect: An estimate of the independent effect of (usually) a factor variable on a response variable in analysis of variance.

Mainframe: High-speed, general-purpose computer with a very large storage capacity.

Majority rule: A requirement that the majority of a series of diagnostic tests are positive before declaring that a patient has a particular complaint. See also unanimity rule.

Malthusian parameter: The rate of increase that a population would ultimately attain if its age-specific birth rate and age-specific death rate were to continue indefinitely. See also population growth model. [Proceedings of the National Academy of Sciences of the United States of America, 1996, 93, 15276–8.]

Management trial: Synonymous with pragmatic trial. Manifest variable: A variable that can be measured directly, in contrast to a latent variable. For example, blood pressure, weight, height, etc.

Calendarization: A generic term for benchmarking.

Calendar plot: A method of describing compliance for individual patients in a clinical trial, where the number of tablets taken per day are set in a calendar-like form (see Figure 12). See also chronology plot. [Statistics in Medicine, 1997, 16, 1653–64.]

Calibration: A procedure that enables a series of easily obtainable but possibly less precise measurements to be used in place of more expensive or more-difficult-to-obtain measurements of some quantity of interest. Suppose, for example, that there is a well-established, accurate method of measuring the concentration of a given chemical compound, but that it is too expensive and/or cumbersome for routine use. A cheap and easy-to-apply alternative is developed that is, however, known to be imprecise and possibly subject to bias. By using both methods over a range of concentrations of the compound, and applying regression analysis to the values from the cheap method and the corresponding values from the accurate method, a calibration curve can be constructed that may, in future applications, be used to read off estimates of the required concentration from the values given by the less involved, inaccurate method. [International Statistical Institute, 1991, 59, 309–36.]

Calibration curve: See calibration. California score: A score used in studies of sudden infant death syndrome that gives the number from eight adverse conditions present for a given infant. The events include fewer than 11 antenatal visits, male sex, birthweight under 3000 g and mother under 25 years old.

Radical Statistics Group: A national network of social scientists in the UK committed to a critique of statistics as used in the policymaking process. The group attempts to build the competence of critical citizens in areas such as health and education. [Radical Statistics Group, 10 Ruskin Avenue, Bradford, UK.]

Radioimmunoassay: An assay performed in clinical and biomedical research laboratories to estimate the concentration of an antigen in a biological specimen. [Clinical Chemistry, 1977, 23, 16247.]

Random: Governed by chance. Not determined completely by other factors. Nondeterministic.

Random allocation: A method for forming treatment and control groups, particularly in the context of a clinical trial. Subjects receive the active treatment or placebo on the basis of the outcome of a chance event, for example tossing a coin. The method provides an impartial procedure for allocation of treatments to individuals, free from personal biases, and ensures a firm footing for the application of significance tests and most of the rest of the statistical methodology likely to be used. Additionally, the method distributes the effects of covariates, both observed and unobserved, in a statistically acceptable fashion. See also block randomization, minimization and biased coin method. [Everitt, B. S. and Pickles, A., 2000, Statistical Aspects of the Design and Analysis of Clinical Trials, Imperial College Press, London.]

QOL: Acronym for quality of life.

Quality-adjusted life-years: An adjustment of life expectancy that reduces the overall expectancy by amounts that reflect the existence of chronic conditions causing impairment, disability and handicap as assessed from health survey data. Useful in economic analyses as a measure of health outcomes that reflects both lives saved and patient's valuation of quality of life in alternative states. [Quality of Life Research, 2002, 11, 37–45.]

Quality-adjusted survival analysis: A methodology for evaluating the effects of treatment and other covariates on survival times that allows consideration of both quality and quantity of life. [Annals of Hematology, 2005, 84, 47–55.]

Quality-adjusted survival times: The weighted sum of different time episodes making up a patient's survival time, with the weights reflecting the quality of life of each period. [Statistics in Medicine, 1993, 12, 975–88.]

Quality assurance: Any procedure or method for collecting, processing or analysing data that is aimed at maintaining or improving the reliability or validity of the data.

Quality control procedures: Statistical procedures designed to ensure that the precision and accuracy of, for example, a laboratory test are maintained within acceptable limits. The simplest such procedure involves a chart (usually called a control chart) with three horizontal lines, one drawn at the target level of the relevant control statistic, and the others, called action lines, drawn at some prespecified distance above and below the target level.

Early detection programme: Synonymous with screening studies.

Early warning system: A term used in disease surveillance for any procedure designed to detect as early as possible any departure from usual or normally observed frequency of phenomena. For example, in developing countries, a change in children's average weights is an early warning signal of nutritional deficiency. [Canadian Medical Association, 2002, 166, 1–2.]

EBM: Abbreviation for evidence-based medicine. Ecological fallacy: A term used when spatially aggregated data are analysed and the results assumed to apply to relationships at the individual level. In most cases, analyses based on area-level means give conclusions very different from those that would be obtained from an analysis of unit-level data. An example from the literature is a correlation coefficient of 0.11 between illiteracy and being foreign-born calculated from person-level data in the USA, compared with a value of ―0.53 between percentage illiteracy and percentage foreign-born calculated from summary state summary statistics. [Statistics in Medicine, 1992, 11, 1209–24.]

Ecological statistics: Procedures for studying the dynamics of natural communities and their relation to environmental variables. [Gotelli, N. J. and Ellison, A. M., 2004, A Primer of Ecological Statistics, Sinauer Associates Inc.]

Ecological study: A study in which the units of analysis are populations or groups of individuals rather than individuals. Used widely in epidemiology, despite their methodological limitations (see ecological fallacy), because of their low cost and convenience. [American Journal of Public Health, 1982, 72, 1336–44.]

Yates's contingency correction: When testing for independence in a contingency table, a continuous probability distribution, namely the chi-squared distribution, is used as an approximation to the discrete probability of observed frequencies, namely the multinomial distribution. To improve this approximation, Yates suggested a correction that involves subtracting 0.5 from the positive discrepancies (observed―expected) and adding 0.5 to the negative discrepancies before these values are squared in the calculation of the usual chi-squared test. If the sample size is larger, then the correction will have little effect on the value of the test statistic. Now no longer needed since exact tests can be used. [Everitt, B. S., 1992, The Analysis of Contingency Tables, 2nd edn, Chapman and Hall/CRC, Boca Raton, FL.]

Yea-saying: Synonym for acquiescence bias.

Years of potential life lost: A measure of the total number of life years lost due to premature death in a population from a certain cause. Premature death is usually defined as death earlier than some given value, for example, 65 or 75 years, or less than the appropriate average life expectancy. An indicator of premature mortality that gives more importance to the causes of death that occurred at younger ages than those that occurred at older ages. [British Medical Journal, 1990, 301, 429–32.]

Abortion rate: The annual number of abortions per 1000 women of reproductive age (usually defined as age 15–44 years). For example, in the USA in 1970 the rate was five, in 1980 it was 25 and in 1990 it was 24. [Family Planning Perspectives, 1998, 30, 244–7.]

Abortion ratio: The estimated number of abortions per 1000 live births in a given year. For example, in the USA in 1970 the ratio was 52, in 1980 it was 359 and in 1990 it was 344. [Family Planning Perspectives, 1998, 30, 244–7.]

Abscissa: The horizontal (or x-axis) on a graph, or a particular point on that axis.

Absolute cause-specific risk: Synonym for absolute risk.

Absolute deviation: Synonym for average deviation.

Absolute risk: Often used as a synonym for incidence, although also used occasionally for attributable risk, excess risk or risk difference. Defined more properly as the probability that a disease-free individual will develop a given disease over a specified time interval given current age and individual risk factors, and in the presence of competing risks. Absolute risk is a probability and consequently lies between 0 and 1. See also relative risk. [Kleinbaum, D.G., Kupper, L. L. and Morgenstern, H., 1982, Epidemiologic Research: Principles and Quantitative Methods, Lifetime Learning Publications, Belmont.]

Absolute risk reduction: The proportion of untreated people who experience an adverse event minus the proportion of treated people who experience the event.