Analysing multiple variables per subject

This chapter describes the statistical issues involved in analysing studies with more than one variable for each subject such as when adjusting for confounding or nuisance variables or wishing to disentangle the effect of multiple variables on a single outcome. There is a wide range of modelling techniques that can be used and so the approaches commonly used when analysing different outcome variables and/or different study designs are described. The chapter discusses the approaches in terms of when the methods are appropriate, how they work, and how the results are interpreted.

Glossary of terms

This chapter contains a glossary of key terms used in the book, in alphabetical order.

Other statistical methods

This chapter describes several individual statistical methods that do not fit neatly in the other chapters but which are commonly used in medical research. These include methods used to assess agreement in measurement and reliability studies, the number needed to treat as a measure of efficacy in a trial, life tables, missing data, and analysing cost data. All methods are illustrated with examples.

Probability and distributions

Probability and probability distributions play a central part in medical statistics. This chapter defines what is meant by probability and describes the rules by which probabilities are combined. It then describes how the use of probability leads to the concept of a probability distribution and shows how these distributions are used in medical statistics. Examples are given of the use of key distributions: the Normal distribution, the binomial distribution, and the Poisson distribution.

Collecting and handling data

This chapter gives suggestions for designing questions and questionnaires, and discusses the consequences of different question designs on the resulting statistical analyses. It also discusses different methods of data entry and handling datasets in computer packages. The importance of checking for errors in the data is highlighted with suggestions of how to do this. Finally, the role of data monitoring committees in research trials is discussed, along with the implications of ending trials early. Examples are provided throughout the chapter.

Analysing multiple observations per subject

This chapter describes the statistical issues involved in analysing studies with more than one data point or observation per subject, such as when a series of measurements are made on each individual over time or when a group of subjects are analysed together, forming a cluster. For each of these situations, the statistical analysis needs to take account of the design of the study, and for most situations there are several possible approaches which may be used. The most common approaches are described in terms of when the methods are appropriate, how they work, and how the results are interpreted.

Presenting research findings

This chapter discusses the different formats for disseminating research findings and the different sections of a research paper or report. It also describes the best ways to present statistical results and discusses guidelines for research articles. Examples are given throughout.

Meta-analysis

This chapter describes the statistical issues involved in performing meta-analyses. It discusses the sources and effects of publication bias and considers ways of correcting for it. It also discusses statistical and clinical heterogeneity and considers how these can be addressed in meta-analyses. Finally, it describes individual patient meta-analysis. Throughout, the chapter includes both trials and observational studies, discussing the challenges that each study design brings and giving examples.

Diagnostic studies

This chapter describes how statistical methods are used in diagnostic testing to obtain different measures of a test’s performance. It describes how to calculate sensitivity, specificity, and positive and negative predictive values, and shows the relevance of the pre- and post-test odds and the likelihood ratio in evaluating a test in clinical practice. The chapter also describes the receiver operating characteristic curve and shows how this links with logistic regression analysis. All methods are illustrated with examples.

Summarizing data

This chapter describes types of quantitative and categorical data and shows how these different types of data can be summarized numerically and in graphs. It gives worked examples of how to calculate the mean, median, standard deviation, and interquartile range, and gives examples of displaying data in graphs.