You are here

Practical Biostatistics: A User-Friendly Approach for Evidence-Based Medicine

Mendel Suchmacher and Mauro Geller
Publisher: 
Academic Press
Publication Date: 
2012
Number of Pages: 
236
Format: 
Paperback
Price: 
69.95
ISBN: 
9780124157941
Category: 
Textbook
[Reviewed by
Russell Jay Hendel
, on
12/31/2012
]

Practical Biostatistics is a clearly written and well laid out textbook. I would recommend its use for either an introductory biostatistics course or a graduate course targeted to students specializing in health-care. This book could also be used in a standard first year undergraduate introductory statistics course if the instructor doesn’t mind overemphasizing clinical applications. If the book is used as a standalone text the instructor may find it helpful to supplement the exercises; but these days there are many available data sets available online.

The book has the following useful features:

Lean and lively look: A comprehensive statistics book with only 236 pages! The text has 46 sections which neatly adapt to a traditional fifteen week course meeting three days a week. About 25% of these sections are “appendices” showing how to implement important statistical tests in Excel. Thus this book could equally serve as a text for an introductory bio-statistics course with a one-day a week lab component.

Software friendly: The book is written assuming Excel as an accompanying software package. The chief virtue of this is that Excel is almost universally available as part of the Microsoft Office suite. All software algorithms are completely explained in the text (this will be detailed below).

Before continuing with the book’s other useful features we should compare Excel with Minitab and SAS two other popular statistical software packages.

  • SAS is definitely more powerful than Excel (or Minitab). However, SAS’s power resides in the number of things it implements; for example, it has over a 100 methods to study outliers. This is useful if you know the mathematics behind these 100 methods. However, if you don’t know the theory behind them (or, more likely, you have never heard of them) SAS is not much different from Excel or Minitab.
  • Minitab is a popular software package extensively used in teaching statistics. Its one chief drawback is accessibility. For example, students who work all day and take evening courses will typically not have Minitab available to do homework. Also, unless the statistics course has a lab component with computer stations and Minitab, the software cannot be used in the course.
  • The virtue of Excel is accessibility; everyone has it or can obtain it easily and cheaply. Excel allows the full gamut of introductory statistical algorithms and has a neat graphical display. Minitab is slightly friendlier than Excel; for example, the Minitab command display descriptive statistics gives you in one click about a dozen basic statistical measures (average, median, etc.). Excel can display each of these items but the drawback is that one has to invoke each display separately; this is not a serious drawback. Furthermore, an instructor familiar with Excel macros can mimic the Minitab single button approach with a macro (that yields the dozen or so basic statistical measures)

Returning to our enumeration of virtues of this book:

Completely detailed and diagrammed algorithms: The text explains every important algorithm and subalgorithm with detailed steps, each step typically being accompanied by a data table and excel snapshot.

The following example is illustrative: determination whether a data distribution is normal or not: Such a determination is an important prior step in implementing certain statistical tests (such as linear regression). Many books only mention this requirement but then do not indicate how to assess normality. Certain texts may even mention tests, such as the Anderson-Darling test, with only a mention of what p-values indicate normality without any intuitive explanation of the test. Other texts may simply mention that you graph the standardized scores and see if they look normal without any further criteria on what looking normal means.

Practical Biostatistics devotes a whole step, step 5 of the investigator hypothesis examination, to test for normality. Step 5 occupies an entire chapter, chapter 8, of the book. Here is how the book approaches this important test.

First: The algorithm of converting data to a curve of standardized scores is presented.

Second: The authors illustrate this method with a detailed data example. Three tables are presented in their entirety — (i) original data, (ii) standardized data, (iii) frequency table of standardized data.

Third: The authors give four criteria (which this reviewer has never seen in a textbook) for assessment of whether the resulting curve is normal: (i) are mean, median and mode close, (ii) is the curve symmetric without skewness or kurtosis (iii) does 95% of the curve lie within two standard deviations of the mean and (iv) is the coefficient of variation between 25% and 50%?

Fourth: The authors then walk the student through going from a data table to a normal curve in Microsoft Excel. Their explanation uses nine Excel snapshots.

Such an attention to detail is important in an introductory statistics course where the students are typically not math majors and need such walkthroughs.

Knowledgeable authors: Both authors are professors of immunology and work as consultants for the pharmaceutical industry in the clinical research field.

Detailed discussion of applications: How refreshing to see Parts I and II of the book — (2 1/2 weeks of lectures) — devoted to a detailed discussion of data studies. Instead of diving right into the math or have a small introductory section discussing the difference between categorical and interval variables, the authors carefully discuss the various types of clinical studies: ecological, cross-sectional, longitudinal, observational, intervention, case-control, cohort, stratified etc. Such discussions make the subject appear real. Such discussions are especially useful for beginning graduate students in the health care field.

User friendly summaries, diagrams and tables: The book is copyrighted 2012 and indeed has the flavor of a post-2010 book. The text is replete with summaries, diagrams and tables. For example the first 11 pages discussing the types of epidemiological studies, has 4 diagrams and an entire section summarizing the various types and subtypes of epidemiological studies. Such user friendly aids are especially useful to undergraduates who may not yet know how to study technical material.

Suggested Readings: It is common these days for books to have bibliographic references to other very useful books and articles. Practical Biostatistics gives URLs of websites with clinical research articles, such as medline, as well as other useful journal articles. For example, I was delighted to find a four-part article series in the CMAJ, Canadian Medical Association Journal, titled “Basic Statistics for Clinicians,” which reviews illustrations of basic statistics needed by clinicians.

Problems: A nice feature of the book is that it presents annotated solutions to all problems (not just the odd number problems).

My only criticism of the book is the lack of sufficient problems. But that is not insurmountable in this day and age, when data sets on a variety of topics are freely available on the web. An instructor can use them for homework, tests and classroom illustration. My favorite compendium of such free data sets is provided by the course home pages of the AP (Advanced Placement) section of the College Board website, accessible at http://apcentral.collegeboard.com/apc/members/courses/teachers_corner/22027.html.

If I had to chose between a text (i) with many homework exercises but a lack of illustrative algorithmic diagrams and a lack of in-depth discussion on types of statistical studies vs. (ii) a text with many illustrative algorithmic diagrams and in-depth discussion on types of statistical studies I would certainly chose (ii) over (i). I would be choosing between providing extra homework problems or providing illustrations of extra steps and it is certainly easier to provide extra homework problems.


Russell Jay Hendel (RHendel@Towson.Edu) holds a Ph.D. in theoretical mathematics and an Associateship from the Society of Actuaries. He teaches at Towson University. His interests include discrete number theory, applications of technology to education, problem writing, actuarial science and the interaction between mathematics, art and poetry.

PART 1 - STUDY TYPE DETERMINATION
1. Epidemiological studies
2. Analytical studies
3. Summary


PART 2 - OBSERVATIONAL STUDIES
4. Determination of association strength between an exposure factor and an event in observational studies
5. Increasing accuracy in observational studies


PART 3 - STEP-BY-STEP BIOSTATISTICS OF A CLINICAL TRIAL
6. STEP 1: Investigator's hypothesis and expression of its corresponding outcome
7. STEP 2: n estimation and n assessment of a published trial
8. STEP 3: Organization of variables and endpoints
9. STEP 4: Measures for results expression of a clinical trial
10. STEP 5: Determination of normality or non-normality of data distribution
11. STEP 6: Hypothesis testing
12. STEP 7: Correlating sample data with the general population - 95% confidence Interval


PART 4 - ADDITIONAL CONCEPTS IN BIOSTATISTICS
13. Individual and collective benefit and risk indexes inferable from intervention cohort studies
14. Statistical assessment of diagnostic tests for the clinic
15. Systematic reviews and meta-analyses
16. Correlation and regression
17. Per protocol analysis and intention to treat analysis


Self-evaluation
Annotated answers
Microsoft Excel 2010 Basics
References
Glossary
Index