Principles of Medical Statistics - A. Feinstein (CRC, 2002) WW.pdf - Bio Med 22 - Aivree

PRINCIPLES OF

MEDICAL

Alvan R. Feinstein, M.D.

CHAPMAN & HALL/CRC

A CRC Press Company

Boca Raton London New York Washington, D.C.

STATISTICS

Library of Congress Cataloging-in-Publication Data

Feinstein, Alvan R., 1925–

Principles of medical statistics / Alvan R. Feinstein.

p. ; cm.

Includes bibliographical references and index.

ISBN 1-58488-216-6 (alk. paper)

1. Medicine—Statistical methods.

[DNLM: 1. Statistics—methods. 2. Data Interpretation,

Statistical. WA 950 F299p 2001] I. Title.

R853.S7 F45 2001

610'.7'27—dc21

2001001794

This book contains information obtained from authentic and highly regarded sources. Reprinted material is quoted with

permission, and sources are indicated. A wide variety of references are listed. Reasonable efforts have been made to publish

reliable data and information, but the author and the publisher cannot assume responsibility for the validity of all materials

or for the consequences of their use.

Neither this book nor any part may be reproduced or transmitted in any form or by any means, electronic or mechanical,

including photocopying, microfilming, and recording, or by any information storage or retrieval system, without prior

permission in writing from the publisher.

The consent of CRC Press LLC does not extend to copying for general distribution, for promotion, for creating new works,

or for resale. Specific permission must be obtained in writing from CRC Press LLC for such copying.

Direct all inquiries to CRC Press LLC, 2000 N.W. Corporate Blvd., Boca Raton, Florida 33431.

Trademark Notice: Product or corporate names may be trademarks or registered trademarks, and are used only for

identification and explanation, without intent to infringe.

Visit the CRC Press Web site at www.crcpress.com

No claim to original U.S. Government works

International Standard Book Number 1-58488-216-6

Library of Congress Card Number 2001001794

Printed in the United States of America 1 2 3 4 5 6 7 8 9 0

Printed on acid-free paper

Preface

What! Yet another book on medical biostatistics! Why? What for?

The purpose of this preface is to answer those questions and to add a few other pertinent

remarks. The sections that follow describe a series of distinctions, some of them unique,

that make this book different from other texts.

Goals and Objectives

The goal of the text is to get biomedical readers to think about data and statistical

procedures, rather than learn a set of “cook-book recipes.” In many statistics books aimed

at medical students or biomedical researchers, the readers are believed to have either

little interest or limited attention. They are then offered a simple, superficial account of

the most common doctrines and applications of statistical theory. The “get-it-over-with-

quickly” approach has been encouraged and often necessitated by the short time given

to statistics in modern biomedical education. The curriculum is supposed to provide

fundamental background for the later careers of medical and other graduate students,

but the heavily stressed “basic science” topics are usually cellular and molecular biology.

If included at all, statistics is usually presented briefly, as a drudgery to be endured

mainly because pertinent questions may appear in subsequent examinations for licensure

or other certifications.

Nevertheless, in later professional activities, practicing clinicians and biomedical

researchers will constantly be confronted with reports containing statistical expressions

and analyses. The practitioners will regularly see and use statistical results when making

clinical decisions in patient care; and the researchers will regularly be challenged by

statistical methods when planning investigations and appraising data. For these activities,

readers who respect their own intellects, and who want to understand and interpret the

statistical procedures, cannot be merely passive learners and compliant appliers of doc-

trinaire customs. The readers should think about what they want, need, and receive. They

should also recognize that their knowledge of the substantive biomedical phenomena is

a major strength and dominant factor in determining how to get, organize, and evaluate

the data. This book is aimed at stimulating and contributing to those thoughts.

Another distinction of the text is that the author is a physician with intimate and

extensive experience in both patient care and biomedical investigation. I had obtained a

master's degree in mathematics before entering medical school, but thereafter my roots

were firmly and irrevocably grounded in clinical medicine. When I later began doing

clinical research and encountering statistical strategies, my old mathematical background

saved me from being intimidated by established theories and dogmas. Although not all

statisticians will approve the temerity of an “unauthorized” writer who dares to compose

a text in which the fundamental basis of old statistical traditions is sometimes questioned,

other statisticians may be happy to know more about the substantive issues contained in

biomedical research, to learn what their clients are (or should be) thinking about, and to

lead or collaborate in developing the new methods that are sometimes needed.

New Methods and Approaches

The text contains many new methods and approaches that have been made possible by

advances in statistical strategy for both analytic description and inferential decisions.

Statistical description has traditionally relied on certain mathematical models, such as

the Gaussian distribution of a “normal” curve, that summarize data with means, standard

deviations, and arbitrarily constructed histograms. Readers who begin to think about what

they really want, however, may no longer happily accept what is offered by those old

models. For example, because biomedical data seldom have a Guassian distribution, the

median is usually a much better summary value than the mean ; and new forms of data

display — the stem-and-leaf plot and the box plot — not only are superior to histograms,

but are more natural forms of expression.

Another descriptive distinction, which is omitted or blurred in many text books, is the

difference between a trend (for citing correlation or regression) and a concordance (for

citing agreement). Investigators who study variability in observers or in laboratory pro-

cedures have usually been taught to express results with the conventional indexes of

“association” that denote trend, but not concordance. This text emphasizes the difference

between correlation and agreement; and separate chapters are devoted to both “nondi-

rectional” concordance (for observer variability) and “directional” concordance (for accu-

racy of marker tests).

In statistical inference for decisions about probability, the customary approach has used

hard-to-understand mathematical theories and hypothetical assumptions that were devel-

oped, established, and entrenched (for topics such as t tests and chi-square tests), because

they led to standard formulas for relatively simple calculations. During the past few

decades, however, the elaborate mathematical theories and assumptions have been aug-

mented, and sometimes replaced, by easy-to-understand new methods, which use rear-

rangements or resamplings of the observed data. The new methods often require

formidable calculations that were not practical in the pre-computer era; but today, the

“computer-intensive” work can be done quickly and easily, requiring no more effort than

pushing the right “button” for an appropriate program. The new methods, which may

eventually replace the old ones, are discussed here as additional procedures that involve

no complicated mathematical backgrounds or unrealistic assumptions about “parametric”

sampling from a theoretical population. In the new methods — which have such names as

Fisher exact test , bootstrap , and jackknife — all of the rearrangements, resamplings, and

statistical decisions about probability come directly from the empirical real-world data.

Another departure from tradition is a reappraisal of the use of probability itself, with

discussions of what a reader really wants to know, which is stability of the numbers, not

just probabilistic assessments.

The text also has sections that encourage methods of “physical diagnosis” to examine

the data with procedures using only common sense and in-the-head-without-a-calculator

appraisals. From appropriate summary statistics and such graphic tactics as box-plot

displays, a reader can promptly see what is in the data and can then make some simple,

effective, mental calculations. The results will often offer a crude but powerful check on

more complex mathematical computations.

A particularly novel and valuable approach is the careful dissection (and proposed

elimination) of the term statistical significance , which has been a source of major confusion

and intellectual pathogenicity throughout 20th-century science. Statistical significance is an

ambiguous term, because it does not distinguish between the theoretical stochastic signif -

icance of calculated probabilities (expressed as P values and confidence intervals) and the

pragmatic quantitative significance or clinical importance of the “effect sizes” found in

the observed results. Not only is the crucial difference between stochastic and quantitative

significance emphasized and thoroughly discussed, but also a special chapter, absent from

conventional texts, is devoted to the indexes of contrast used for expressing and evaluating

the “effect size” of quantitative distinctions.

Two other unique features of this text are the following:

• Two chapters on the display of statistical data in tables, charts, and graphs

contain good and bad examples that can be helpful to readers, investigators, and

the artists who prepare medical illustrations.

• A chapter that discusses the challenges of evaluating “equivalence” rather than

“superiority” also considers the management of problems that arise when dis-

cordance arises in what the investigator wants, what the results show, and what

the statistical tests produce.

Sequence, Scope, Rigor, and Orientation

The text is arranged in a logical sequence of basic principles that advance from simple to

more elaborate activities. It moves from evaluating one group of data to comparing two

groups and then associating two variables. Thereafter, the scope extends into more com-

plex but important topics that frequently appear as challenges in biomedical literature:

controversies about stochastic issues in choosing one- or two-tailed tests, the graphic

patterns of survival analysis, and the problems of appraising “power,” determining

“equivalence,” and adjudicating “multiple hypotheses.”

Nevertheless, despite some of the cited deviations from customary biostatistical dis-

course, the text describes all the conventional statistical procedures and offers reasonably

rigorous accounts of many of their mathematical justifications. Whether retaining or reject-

ing the conventional procedures, a reader should know what they do, how they do it, and

why they have been chosen to do it. Besides, the conventional procedures will continue

to appear in biomedical literature for many years. Learning the mechanisms (and limita-

tions) of the traditional tactics will be an enlightened act of self-defense.

Finally, although the conventional mathematical principles are given a respectful

account, the book has a distinctly clinical orientation. The literary style is aimed at bio-

medical readers; and the examples and teaching exercises all come from the real-world

medical phenomena. The readers are not expected to become statisticians, although appro-

priate historical events are sometimes cited and occasional mathematical challenges are

sometimes offered. Clinical and biomedical investigators have made many contributions

to other “basic” domains, such as cell and molecular biology, and should not be discour-

aged from helping the development of another “basic” domain, particularly the bio -portion

Principles of Medical Statistics - A. Feinstein (CRC, 2002) WW.pdf

Plik z chomika:

Inne pliki z tego folderu:

Inne foldery tego chomika: