Book Reviews
Primer of Biostatistics, 5th Edition
Stanton A. Glantz
New York: McGraw-Hill, 2002,
489 pp., illus., index, $34.95
ISBN: 0-07-137946-0
For nearly 70 years, statisticians have struggled to help scientists
understand statistical concepts and use statistical methods. A book
that promotes literacy in statistics would be a useful resource for
researchers. This 5th edition of Primer of Biostatistics is not that
book.
This Primer does have constructive elements, but they are overshadowed
by problems with the conceptual origin of the book, by information
that is confusing, misleading, or just plain wrong, and by a tone that
sounds sarcastic and all-knowing.
In the preface, the author discusses the conceptual origin of the book
and its impact on the content of the early chapters. This paragraph,
taken from page xvi of the preface, illustrates my concerns with the
book:
“[The book] is based on the premise that much of what is published in
the biomedical literature uses dubious statistical practices, … . Most
of the errors (at least as they relate to statistical inference)
center on misuse of the t test, probably because the people doing the
research were unfamiliar with anything else. … Since so much is
published that should probably be analyzed with analysis of variance,
and since analysis of variance is really the paradigm of all
parametric statistical tests, I present it first, then discuss the t
test as a special case.”
If the conceptual origin of a book is to be a pet peeve—the failure to
use analysis of variance when it is appropriate—the peeve must make
sense from the perspectives of context and logic. In this case, it
does not. Moreover, I am confused by the statement that analysis of
variance is really the paradigm of all parametric statistical tests.
Does the author mean that analysis of variance is the procedure on
which all other parametric tests are based? It is not: analysis of
variance is just one kind of general linear model. But what will an
inexperienced reader think?
The author’s use of analysis of variance as the conceptual origin of
his book creates immediate problems. In Chapter 3, he uses analysis of
variance to compare cardiac output among four groups and then misuses
analysis of variance to compare glucose levels between two groups
(page 47). I raised my eyebrows, too: a two-sample t test would have
been the appropriate procedure. In Chapter 4 is a section entitled
“The t Test is an Analysis of Variance” (page 84). It is not: a
t test
and an analysis of variance do produce equivalent results, but the
structure of the procedures is quite different.
For reasons I fail to understand, many books of statistics depict t
distributions that are inaccurate and inconsistent. So does this one.
The shape of the t distributions in Figures 4-5, 6-4, 6-5, and 6-7 is
nothing like it should be (Figure 1):
Figure 1.
What will happen when an inexperienced reader sees the
inaccurate and inconsistent t distributions?
The Primer stumbles also on simple concepts. For example, the author
reinforces misunderstanding of the adjective null when he states that
the null hypothesis is one of no effect (page 31). This is a common
misconception. The null hypothesis is the hypothesis being tested; it
need not be one of no difference.
At the beginning of this review, I mentioned that the Primer does have
constructive elements. What are they? The discussion that standard
deviations rather than standard errors estimate variability; the
discussion that statistical significance and scientific importance
differ; the discussion that confidence intervals are important. These
elements, however, are swamped by problems that are bigger.
In the concluding chapter (page 436), the author asks how the reader
can help improve the use of statistics by researchers. He answers by
writing:
“Do not let people get away with sloppy statistical thinking any more
than you would permit them to get away with sloppy clinical or
scientific thinking. Write letters to the editor. Ask questions in
class, rounds, and meetings.”
Whatever happened to the notion of collaboration between scientist and
statistician as a way to improve literacy in statistics?
Because of this omission, because of problems with
the conceptual origin of the book, and because of information that is
confusing or misleading, I am unable to recommend Primer of
Biostatistics as a resource.
Douglas Curran-Everett
National Jewish Medical
and Research Center
University of Colorado
Health Sciences Center
Michael J. Brownstein and Arkady B. Khodursky (Editors).
Totowa, NJ: Humana Press, 2003,
258 pp., $89.50
ISBN: 1-588-29291-6
Microarray technology is a major experimental platform that has become a mainstay in the realm of functional genomic explorations of a wide range of biological systems. As with any scientific experiment, there are essentially four parts that constitute a microarray experiment: 1) designing the experiment; 2) conducting the experiment, 3) analyzing the data obtained, and 4) interpreting the data obtained. Due to the massive quantity of data that this technique produces, careful considerations to experimental design issues go a long way in discerning the outcome of a microarray experiment. Skills at conducting microarray experiments are as important as data normalization, analysis, and interpretation. Due to these multiple levels of complexities associated with the performance of microarray experiments, a good one-stop reading material would benefit entry-level as well as senior investigators into the field of microarrays. Functional Genomics: Methods and Protocols edited by Brownstein and Khodursky meets this need very well. This book belongs to the series of Methods in Molecular Biology and is basically comprised of two sections: 1) methods in microarray data generation and 2) methods in microarray data analysis. There is nothing comparable to knowledge gained methods in data generation. Investigators who choose to use this book as a guide to perform microarray experiments will like the attention to detail regarding issues as simple as the choice of tubes to hold reaction mixtures. Time-tested detailed protocols with useful little “tricks” such as, for example, “tap the tube gently to ensure that pellets are dislodged” are written to ensure that readers get the most out of using the recommended protocols in their labs. In our opinion, this is the most valuable section of this book. The chapters on isolation of polysomal RNA for microarray analysis and parallel analysis of gene copy number and expression using cDNA microarrays contain protocols for extended applications of the array technology beyond merely knowing what genes are differentially expressed between two given biological samples. Another particularly interesting chapter is the genome-wide mapping of protein-DNA interactions by chromatin immunoprecipitation and DNA microarray hybridization. This chapter is explained with protocols applied to understand protein-DNA interactions in yeast.
The second section of this book concentrates on experimental design and analysis issues. The organization of experimental design after the section on experimental methods at the bench is somewhat peculiar considering that, in practice, serious considerations to experimental design should be given prior to actually performing the experiments. In any case, topics covered in these chapters are written by well known groups of investigators in the field and include: 1) how to design an informative microarray experiment, 2) descriptions of statistical methods for data normalization, and 3) subsequent analyses for extracting meaningful data from the several sources of experimental noise that emanate out of gene expression experiments. Real examples expose the strengths and weaknesses of each method for a given situation, aimed at helping readers choose appropriate protocols and utilize them for their own data sets. In addition, web links are provided to the programs and tools discussed in several chapters. This book is an excellent reference not only for core bioinformatics/genomics courses in undergraduate and graduate programs but also for researchers involved in setting up core laboratories for microarrays. It is valuable because it includes two additional aspects: guidance for the manufacture and use of spotted microarrays on glass, plastic, and nylon membranes, and a chapter dedicated for management of microarray databases that is crucial for successful data mining long after the actual hybridization experiments are done.
All said and done, this is a useful book to own as a guide for adopting protocols for gene expression analysis.
Bina Joe and Steven L. Britton
Medical College of Ohio
Books Received
E.S.E. Hafez with B. Hafez and S.D. Hafez.
The Encyclopedia of Visual Medicine Series.
Boca Raton, FL: CRC Press, 2004, 250 pp., illus, index, $149.95.
ISBN: 1-84214-235-6.
Elastomeric Proteins: Structures, Bio-mechanical Properties, and Biological Roles
Peter R. Shewry, Arthur S. Tatham, and Allen J. Bailey (Editors).
New York: Cambridge Univ. Press, 2003, 391 pp., illus., index, $95.00.
ISBN: 0-521-81594-0.
Evidence–Based Practice Manual: Research and Outcome Measures in Health and Human Services.
Albert R. Roberts and Kenneth R. Yeager (Editors).
New York: Oxford University Press, 2004, 1050 pp., illus., index, $89.50.
ISBN: 0-19-516500-4.
Microcosms of the Brain.
Douglas Tweed.
New York: Oxford University Press, 2003, 199 pp., illus., index, $37.50.
ISBN: 0-19—852893-0.
Mind Time: The Temporal Factor in Consciousness.
Benjamin Libet.
Cambridge, MA: Harvard University Press, 248 pp., illus., index, $29.95.
ISBN: 0-674-01320-4.