Andrew Blann (2015), Data Handling and Analysis, New York, Oxford University Press, Fundamentals of Biomedical Science, 196pp
Review by Joshua Hardy, Department of Biochemistry and Molecular Biology, Monash University
In this digital age, the generation and storage of data has never been easier. Yet the interpretation of all this information remains critical and the consequences of misinterpretation can be severe. For a biomedical scientist, conclusions drawn can affect the future treatment of patients. Data Handling and Analysis by Andrew Blann is a short textbook which covers the most common methods for statistical analysis and interpretation of experimental and clinical data. It describes the entire experimental process from designing the experiment, to data collection, analysis, and presentation, and finally communicating the findings to fellow scientists.
Most scientists have some understanding of statistics, such as the importance of controls and blinding. However, Blann is aware that the reader may not have a rigorous understanding of mathematics and chemistry. The book begins with defining different types of data, units of measurement, and some basic chemistry. While this may not be as relevant for clinical scientists, he includes a chapter on auditing that may be more useful. The core of the textbook deals with statistical methods and is the section on which the reader will spend most time. He begins with the analysis of simple binomial data (e.g. male/female) right through to complex methods for large datasets. Since the range of statistical software available today means that most researchers do not perform manual calculations, the textbook focuses on the selection of the appropriate analysis method instead. A common mistake in statistics is confusing normally and non-normally distributed data. Blann has provided worked examples of how such mistakes can lead to invalid conclusions.
There are obvious teaching techniques used throughout the textbook. Each chapter is presented in a topical format and contains figures, worked examples, questions, and a chapter summary. Key terms are defined in the margins and self-check questions ensure the reader has digested the information. At the end of the book is an extended activity which combines several analysis methods discussed in the previous chapters and which serves as an excellent revision tool. Also, various analysis software packages are recommended depending on the complexity of the reader's dataset. While primarily intended as a textbook for an undergraduate biomedical science course, the format of the textbook would also make an excellent quick reference guide for the intermediate researcher.
While the intention of this textbook is to provide a simple introduction to the topic it does not gloss over the challenges of interpreting research data in a broader context. Blann explains that correlation does not equal causation; a statistically significant difference may have several confounding factors. He gives examples of potential pathophysiological interpretations for different experiments to illustrate this point and cautions the reader to avoid inferring causality. He also emphasises the importance of verifying the accuracy of data. For example, mistyped values such as an age of 789 can be easily overlooked in a large dataset but greatly increase the variance of a sample and ultimately change the conclusion of a clinical trial.
Blann has successfully compiled a book that is both accessible to the beginner and useful for the intermediate reader. This is not a textbook for professional statisticians, and by limiting the audience it increases the relevancy for other readers. While similar online resources exist for data analysis, the discipline-specific nature of this book and its concise format make it an excellent resource for the budding researcher. A researcher equipped with the analysis tools Blann provides will be able to critically engage in a field of growing complexity and contribute to the body of knowledge in their discipline. I am 95% certain of that.
Review by Philip Young, School of Life Sciences, University of Warwick
Right from the start, it should be highlighted that this book is not aimed at basic scientists or academics. It is aimed specifically at students and academics who want to become clinical scientists in diagnostic laboratories in the NHS. If you are an undergraduate student who simply wishes to improve your background statistical knowledge, there are better books that will contain more relevant information. This book concentrates on clinically relevant statistics, starting with basic data handling, progressing to assay validation, diagnostic statistics, ANOVA and survival analysis. Although most of these are relevant to biomedical and medical students, there are additional sections on scientific audits, NEQAS and CPA processes that are not.
Students, irrespective of their background, tend to struggle with contextualising their statistical knowledge, even those with good A-Level mathematics grades. This is one of the main strengths of this book; it provides working examples, described in a simple-to-follow and basic way. For example, the section on medical statistics uses a direct comparison between two tests for cancer diagnosis to explain the core differences between sensitivity, specificity, positive predictive value and negative predictive value. This allows students to visualise the central differences, which should improve concept retention by preventing the text becoming 'dry' – a common issue with statistical sources.
However, there are several issues with the text. Firstly, and most importantly, the difference between standard error of the mean (SEM; a measure of precision) and standard deviation (SD; a measure of variance) is not dealt with. The inappropriate use of SEM as a measure of variance is a major issue in scientific analysis and writing, so much so that several leading scientific journals have recently been compelled to publish editorials highlighting why SEM is not a measure of variance. The failure of this book adequately to explain the difference between the two is a major flaw. Secondly, there is no logical flow to the sections. For example, relatively complex topics like Bland-Altman limits of agreement, likelihood ratios and coefficients of variation are dealt with before the very basic section on graphical presentation of data. Although both sections are good, considering Bland-Altman analysis is based on a graphical presentation of the data, the order in which they appear in the text makes little sense.
In all, although there are issues with the topics covered, the book is a good overview of clinical science and the statistics needed for working in a clinical diagnostic lab. However, as initially stated, there are better basic statistics textbooks for biomedical students who wish to enter either medicine or PhD level research.
To cite either of these reviews please use the following details: Hardy, J OR Young, P. (2015), Andrew Blann (2015), 'Data Handling and Analysis', Reinvention: an International Journal of Undergraduate Research, Volume 8, Issue 1, http://www.warwick.ac.uk/reinventionjournal/archive/volume8issue1/hardyyoung Date accessed [insert date]. If you cite these reviews or use them in any teaching or other related activities please let us know by e-mailing us at Reinventionjournal at warwick dot ac dot uk.