Biostatistics for Medical and Biomedical Practitioners (eBook)

Julien I. E. Hoffman (Autor)

eBook Download: PDF | EPUB

2015 | 1. Auflage
770 Seiten
Elsevier Science (Verlag)
978-0-12-802607-6 (ISBN)

Infos

Biostatistics for Practitioners: An Interpretative Guide for Medicine and Biology deals with several aspects of statistics that are indispensable for researchers and students across the biomedical sciences. The book features a step-by-step approach, focusing on standard statistical tests, as well as discussions of the most common errors. The book is based on the author's 40+ years of teaching statistics to medical fellows and biomedical researchers across a wide range of fields. - Discusses how to use the standard statistical tests in the biomedical field, as well as how to make statistical inferences (t test, ANOVA, regression etc.) - Includes non-standards tests, including equivalence or non-inferiority testing, extreme value statistics, cross-over tests, and simple time series procedures such as the runs test and Cusums - Introduces procedures such as multiple regression, Poisson regression, meta-analysis and resampling statistics, and provides references for further studies

Julien I E Hoffman, M.D., F.R.C.P (London) was born and educated in Salisbury (now Harare) in Southern Rhodesia (now Zimbabwe). He received a Bsc (Hons) in 1945 from the University of the Witwatersrand in South Africa, and his M.B., B.Ch. degree there in 1949. After working in the Departments of Medicine in Johannesburg General Hospital and in the Central Middlesex Hospital in London, he worked for the Medical Research Council at the Royal Postgraduate School in Hammersmith, London. Then he spent two years training in Pediatric Cardiogy at Boston Children's Hospital, followed by 15 months as a Fellow at the Cardiovascular Research Institute (CVRI) at the University of California in San Francisco (UCSF).In 1962 he joined the faculty of the Albert Einstein College of Medicine in New York, and moved in 1966 to UCSF as Associate Professor of Pediatrics and member of the CVRI. He spent 50% of his time in the care of children with heart disease and 50% of his time doing research into the pathophysiology of the coronary circulation.His interest in Statistics began while taking his Science degree. In England, he took a short course run by Bradford Hill. On returning to Johannesburg he was assigned to statistical analyses for other members of the Department of Medicine. Learning was by trial and error, helped by Dr J Kerrich, head of the University's Statistics Department. Hoffman began teaching statistics to Medical students in 1964, and in San Francisco conducted an approved course for Fellows and Residents for over 30 years. He was a member of the Biostatistics group for approving and coordinating statistics at UCSF. For many years he was a statistical consultant for the journal Circulation Research, and was intermittently statistical consultant to several other medical journals.

Textauszug

Preface

In my 40 years experience of advising medical research workers about how to analyze their studies, certain problems arose frequently. For examples, many investigators wanted to compare two Poisson distributions, yet some introductory books on Biostatistics give little attention to the Poisson distribution, even though it is an important distribution that answers the question about how often a rare event occurs, such as the number of deliveries per hour in a delivery room. Few people can navigate the minefield of multiple comparisons, involved when several different groups are compared, often done incorrectly by performing multiple t-tests, yet most elementary texts do not deal with this problem adequately. Problems of repeated measures analysis in which several measurements are made in each member of the group and thus are not independent occur frequently in medical research but are not often discussed. Tolerance tests are often needed to set ranges of normal values so that a single measurement can be assessed as likely to be normal or abnormal (such as a single fasting blood glucose concentration). Because most basic books do not discuss this problem, most people incorrectly set confidence limits that should apply only to mean values. In fact, one of the incentives for this book was the lack of books in introductory Biostatistics that could be understood relatively easily, but nevertheless were advanced enough that most investigators would not need to buy additional books or hunt through unfamiliar journals for appropriate tests.

The book is intended to help physicians and biologists who might have had a short course on Statistics several years ago, but have forgotten all but a few of the terms and concepts and have not used their knowledge of statistics for reading the literature critically or for designing experiments. The general aim is to extend their knowledge of statistics, to indicate when various tests are applicable, what their requirements are, and what can happen when they are used inappropriately.

This book has four components.

1. It covers the standard statistical approaches for making descriptions and inferences—for example, mean and standard deviation, confidence limits, hypothesis testing, t-tests, chi-square tests, binomial, Poisson, and normal distributions, analysis of variance, linear regression and correlation, logistic regression, and life tables—to help readers understand how the tests are constructed, how to look for and avoid using inappropriate tests, and how to interpret the results. Examples of injudicious use of these tests are given. Although some basic formulas are presented, these are not essential for understanding what the tests do and how they should be interpreted.

2. Some chapters include a section on advanced methods that should be ignored on a first reading but provide information when needed, and others have an appendix where some simple algebraic proofs are given. As this is not intended to be a mathematically rigorous book, most mathematical proofs are omitted, but a few are important teaching tools in their own right and should be studied. However, knowledge of mathematics (and differential calculus in particular) beyond elementary algebra is not required to use the material provided in this book. The equations indicate what the tests are doing.

3. Scattered throughout the chapters are variations on tests that are often needed but not frequently found in basic texts. These sections are often labeled “Alternative Methods,” and they should be read and understood because they often provide simpler and more effective ways of approaching statistical inference. These include:

a. Robust statistics for dealing with grossly abnormal distributions, both univariate and bivariate.

b. Extending McNemar's test, a test for comparing paired counts, to more than two categories; for example, if matched pairs of patients are given one of two treatments and the results recorded as improved, the same, or worse, how should these be analyzed?

c. Equivalence or noninferiority testing, to determine if a new drug or vaccine is equivalent to or not inferior to those in standard use.

d. Finding the break point between two regression lines. For example, if the lactate:pyruvate ratio remains unchanged when systemic oxygen delivery is reduced below normal until some critical point is reached when the ratio starts to rise, how do we determine the critical oxygen delivery value?

e. Competing risks analysis used when following the survival of a group of patients after some treatment, say replacement of the mitral valve, and allowing for deaths from noncardiac causes.

f. Tolerance testing to determine if a single new measurement is compatible with a normative group.

g. Crossover tests, in which for a group of subjects each person receives two treatments, thus acting as his or her own control.

h. Use of weighted kappa statistics for evaluating how much two observers agree on a diagnosis.

    Some of these analyses can be found only in journals or advanced texts, and collecting them here may save investigators from having to search in unfamiliar sources to find them.

4. Some chapters describe more complex inferences and their associated tests. The average investigator is not likely to use any of these tests without consultation with a statistician, but does need to know that these techniques exist and, if even vaguely, what to look for and how to interpret the results of these tests when they appear in publications. These subjects include:

a. Poisson regression (Chapter 34), in which a predicted count, for example, the number of carious teeth, is determined by how many subjects have zero, 1, 2, etc., carious teeth.

b. Resampling methods (Chapter 37), in which computer-intensive calculations allow the determination of the distributions and confidence limits for mean, median, standard deviations, correlation coefficients, and many other parameters without needing to assume a particular distribution.

c. The negative binomial distribution (Chapter 19) that allows investigation of distributions that are not random but in which the data are aggregated. If we took samples of seawater and counted the plankton in each sample, a random distribution of plankton would allow us to fit a standard distribution such as a binomial or Poisson distribution. If, however, some samples had excessive numbers of plankton and others had very few, a negative binomial distribution may be the way to evaluate the distribution.

d. Meta-analysis (Chapter 36), in which the results of several small studies are aggregated to provide a larger sample, for example, combining several small studies of the effects of beta-adrenergic blockers on the incidence of a second myocardial infarction, is often used. The pitfalls of doing such an analysis are seldom made clear in basic statistics texts.

e. Every investigator should be aware of multiple and nonlinear regression techniques (Chapter 30) because they may be important in planning experiments. They are also used frequently in publications, but usually without mentioning their drawbacks.

With the general availability of personal computers and statistical software, it is no longer necessary to detail computations that should be done by computer programs. There are many simple free online programs that calculate most of the commonly used statistical descriptions (mean, median, standard deviation, skewness, interquartile distance, slope, correlation, etc.) as well as commonly used inferential tests (t-test, chi-square, ANOVA, Poisson probabilities, binomial probabilities, life tables, etc.), along with their associated graphics. More complex tests require commercial programs. There are free online programs for almost all the tests described in this book, and hyperlinks are provided for these.

Problems are given in appropriate chapters. They are placed after a procedure is described so that the reader can immediately practice what has been studied to make sure that the message is understood; the procedures are those that should be able to be performed by the average reader without statistical consultation. Although the problems are simple and could be done by hand, it is better to use one of the recommended online calculators because they save time and do not make arithmetic errors. This frees up time for the reader to consider what the results mean.

The simpler arithmetic techniques, however, are still described in this book because they lead to better understanding of statistical methods, and show the reader where various components of the calculation come from, and how the components are used and interpreted. In place of tedious instructions for doing the more complex arithmetic procedures, there is a greater concentration on the prerequisites for doing each test and for interpreting the results. It is easier than ever for the student to think about what the statistical tests are doing and how they contribute to solving the problem. On the other hand, we need to resist the temptation to give a cookbook approach to solving problems without giving some understanding of their bases, even though this may involve some elementary algebra. As Good and Hardin (2009) wrote: “Don't be too quick to turn on the computer. Bypassing the brain to...

Produkt-Details

• Erscheint lt. Verlag: 3.9.2015
• Sprache: englisch
• Themenwelt: Mathematik / Informatik ► Mathematik ► Angewandte Mathematik
• Themenwelt: Medizin / Pharmazie ► Allgemeines / Lexika
• Themenwelt: Medizin / Pharmazie ► Medizinische Fachgebiete ► Biomedizin
• Themenwelt: Medizin / Pharmazie ► Physiotherapie / Ergotherapie ► Orthopädie
• Themenwelt: Studium ► Querschnittsbereiche ► Epidemiologie / Med. Biometrie
• Themenwelt: Naturwissenschaften ► Biologie
• ISBN-10: 0-12-802607-3 / 0128026073
• ISBN-13: 978-0-12-802607-6 / 9780128026076