Ton J. Cleophas and
Aeilko H. Zwinderman
SPSS for Starters and 2nd Levelers2nd ed. 2016
Ton J. Cleophas
Department Medicine, Albert Schweitzer
Hospital, Dordrecht, The Netherlands
European College Pharmaceutical
Medicine, Lyon, France
Aeilko H. Zwinderman
European College Pharmaceutical
Medicine, Lyon, France
Department Biostatistics, Academic
Medical Center, Amsterdam, The Netherlands
ISBN 978-3-319-20599-1e-ISBN 978-3-319-20600-4
DOI 10.1007/978-3-319-20600-4
Springer Cham Heidelberg New
York Dordrecht London
Library of Congress Control
Number: 2015943499
© Springer International Publishing
Switzerland 2016
Additional material to this book can be
downloaded from http://extras.springer.com
This work is
subject to copyright. All rights are reserved by the Publisher,
whether the whole or part of the material is concerned,
specifically the rights of translation, reprinting, reuse of
illustrations, recitation, broadcasting, reproduction on microfilms
or in any other physical way, and transmission or information
storage and retrieval, electronic adaptation, computer software, or
by similar or dissimilar methodology now known or hereafter
developed.
The use of
general descriptive names, registered names, trademarks, service
marks, etc. in this publication does not imply, even in the absence
of a specific statement, that such names are exempt from the
relevant protective laws and regulations and therefore free for
general use.
The publisher, the
authors and the editors are safe to assume that the advice and
information in this book are believed to be true and accurate at
the date of publication. Neither the publisher nor the authors or
the editors give a warranty, express or implied, with respect to
the material contained herein or for any errors or omissions that
may have been made.
Printed on acid-free
paper
Springer
International Publishing AG Switzerland is part of Springer
Science+Business Media (www.springer.com)
Prefaces to the 1st
edition
Part I
This small book addresses different
kinds of data files, as commonly encountered in clinical research
and their data analysis on SPSS software. Some 15 years ago serious
statistical analyses were conducted by specialist statisticians
using mainframe computers. Nowadays, there is ready access to
statistical computing using personal computers or laptops, and this
practice has changed boundaries between basic statistical methods
that can be conveniently carried out on a pocket calculator and
more advanced statistical methods that can only be executed on a
computer. Clinical researchers currently perform basic statistics
without professional help from a statistician, including t-tests
and chi-square tests. With the help of user-friendly software, the
step from such basic tests to more complex tests has become smaller
and more easy to take.
It is our experience as masters’ and
doctorate class teachers of the European College of Pharmaceutical
Medicine (EC Socrates Project, Lyon, France) that students are
eager to master adequate command of statistical software for that
purpose. However, doing so, albeit easy, it still takes 20–50 steps
from logging in to the final result, and all of these steps have to
be learned in order for the procedures to be successful.
The current book has been made
intentionally small, avoiding theoretical discussions and
highlighting technical details. This means that this book is unable
to explain how certain steps were made and why certain conclusions
were drawn. For that purpose additional study is required, and we
recommend that the textbook “Statistics Applied to Clinical
Trials,” Springer 2009, Dordrecht, Netherlands, by the same
authors, be used for that purpose, because the current text is much
complementary to the text of the textbook.
We have to emphasize that automated
data analysis carries a major risk of fallacies. Computers cannot
think and can only execute commands as given. As an example,
regression analysis usually applies independent and dependent
variables, often interpreted as causal factors and outcome factors.
For example, gender or age may determine the type of operation or
type of surgeon. The type of surgeon does not determine the age and
gender. Yet a software program does not have difficulty to use
nonsense determinants, and the investigator in charge of the
analysis has to decide what is caused by what, because a computer
cannot do things like that, although they are essential to the
analysis. The same is basically true with any statistical tests
assessing the effects of causal factors on health outcomes.
At the completion
of each test as described in this book, a brief clinical
interpretation of the main results is given in order to compensate
for the abundance of technical information. The actual calculations
made by the software are not always required for understanding the
test, but some understanding may be helpful and can also be found
in the above textbook. We hope that the current book is small
enough for those not fond on statistics but fond on statistically
proven hard data in order to start
on SPSS , a software program with an excellent state of the
art for clinical data analysis. Moreover, it is very satisfying to
prove from your own data that your own prior hypothesis was true,
and it is even more satisfying if you are able to produce the very
proof yourself.
Part II
The small book “SPSS for Starters”
issued in 2010 presented 20 chapters of cookbook-like step by step
data analyses of clinical research and was written to help clinical
investigators and medical students analyze their data without the
help of a statistician. The book served its purpose well enough,
since 13,000 electronic reprints were being ordered within 9 months
of the edition.
The above book reviewed, e.g., methods
for:
1.
Continuous data, like t-tests,
nonparametric tests, and analysis of variance
2.
Binary data, like crosstabs, McNemar’s
tests, and odds ratio tests
3.
Regression data
4.
Trend testing
5.
Clustered data
6.
Diagnostic test validation
The current book is a logical
continuation and adds further methods fundamental to clinical data
analysis.
It contains, e.g., methods for:
1.
Multistage analyses
2.
Multivariate analyses
3.
Missing data
4.
Imperfect and distribution free
data
5.
Comparing validities of different
diagnostic tests
6.
More complex regression models
Although a wealth of computationally
intensive statistical methods is currently available, the authors
have taken special care to stick to relatively simple methods,
because they often provide the best power and fewest type I errors
and are adequate to answer most clinical research questions.
It is time for clinicians not to get
nervous anymore with statistics and not to leave their data anymore
to statisticians running them through SAS or SPSS to see if
significances can be found. This is called data dredging.
Statistics can do more for you than produce a host of irrelevant
p-values. It is a discipline at the interface of biology and
mathematics: mathematics is used to answer sound biological
hypotheses. We do hope that “SPSS for Starters 1 and 2” will
benefit this process.
Two other
publications from the same authors entitled Statistical Analysis of Clinical Data on a
Pocket Calculator 1 and 2 are rather complementary to the
above books and provide a more basic approach and better
understanding of the arithmetic.
Ton J. Cleophas
Aeilko H. Zwinderman
Lyon,
France
December 2009, January 2012
Preface to 2nd
edition
Over 100,000 copies of various
chapters of the first edition of SPSS for Starters (Parts I (2010)
and II (2012)) have been sold, and many readers have commented and
given their recommendations for improvements.
In this 2nd edition, all the chapters
have been corrected for textual and arithmetic errors, and they
contain updated versions of the background information, scientific
question information, examples, and conclusions sections. In “notes
section”, updated references helpful to a better understanding of
the brief descriptions in the current text are given.
Instead of the, previously published,
two-20-chapter Springer briefs, one for simple and one for complex
data, this 2nd edition is produced as a single 60-chapter
textbook.
The, previously used, rather arbitrary
classification has been replaced with three parts, according to the
most basic differences in data file characteristics:
1.Continuous outcome data (36
chapters)
2.Binary outcome data (18
chapters)
3.Survival and longitudinal data (6
chapters)
The latter classification should be
helpful to investigators for choosing the appropriate class of
methods for their data.
Each chapter now starts with a
schematic overview of the statistical model to be reviewed,
including types of data (mainly continuous or binary (yes, no)) and
types of variables (mainly outcome and predictor variables).
Entire data tables of the examples are
available through the Internet and are redundant to the current
text. Therefore, the first 10 rows of each data table have now been
printed only.
However, relevant details about the
data have been inserted for improved readability.
Also simple explanatory graphs of the
principles of the various methods applied have been added.
Twenty novel chapters with methods,
particularly, important to clinical research and health care were
still missing in the previous edition, and have been added.
The current edition focuses on the
needs of clinical investigators and other nonmathematical health
professionals, particularly those needs, as expressed by the
commenters on the first edition.
The arithmetic is still more of a
no-more-than high-school level, than that of the first edition,
while complex computations are described in an explanatory
way.
With the help of several new
hypothesized and real data examples, the current book takes care to
provide step-by-step data-analyses of the different statistical
methodologies with improved precision.
Finally, because of lack of time of
this busy group of people, as expressed by some readers, we have
given additional efforts to produce a text as succinct as possible,
with chapters, sometimes, no longer than three pages, each of which
can be studied without the need to consult others.
Ton J. Cleophas
Aeilko H. Zwinderman
Lyon,
France
January 2015
Contents
Part I Continuous
Outcome Data
1 One-Sample Continuous Data (One-Sample T-Test,
One-Sample Wilcoxon Signed Rank Test, 10
Patients) 3
8
Note 6
7
Conclusion 10
8
Note 10
7
Conclusion 15
8
Note 15
7
Conclusion 21
8
Note 21
6
Conclusion 28
7
Note 28
6
Conclusion 33
7
Note 34
8
Conclusion 40
9
Note 40
7
Conclusion 45
8
Note 45
7
Conclusion 50
8
Note 51
6
Conclusion 56
7
Note 57
6
Conclusion 65
7
Note 65
8
Conclusion 72
9
Note 73
7
Conclusion 77
8
Note 78
6
Conclusion 84
7
Note 84
6
Conclusion 87
7
Note 88
8
Conclusion 93
9
Note 93
7
Conclusion 100
8
Note 100
6
Conclusion 107
7
Note 107
10
Conclusion 169
11
Note 169
Part II Binary Outcome
Data
10
Conclusion 272
11
Note 272
10
Conclusion 301
11
Note 301
Part III Survival and
Longitudinal Data
Index373