Intuitive Biostatistics 4th Edition

Intuitive Biostatistics has a unique informal style distinct from other introductory statistics books. Its goal is always to help a scientist understand what statistical analyses mean. In contrast, it often seems that other books often focus on helping a scientist understand the mathematical basis of the statistical analyses.

Intuitive Biostatistics covers  a broader range of topics than most other introductory texts. Here is a list of topics covered by Intuitive Biostatistics (abbreviated IB below) but not by other texts  (the abbreviations refer to book titles tabulated below).   

How intuitive is statistical thinking?

Chapter 1 of IB shows how poorly the human brain handles probability, and how easy it is to be mislead by data. None of the other books have a similar chapter.

Error bars: SD vs. SEM

A common point of confusion is the distinction between error bars that represent standard deviations (SD) and those that represent standard errors of the mean (SEM), and this is the focus of Chapter 14 of IB.  MS and PB also explain the distinction between SD and SEM (but without as much emphasis as IB). BF, BA and SL define both terms, but provide no guidance to prevent confusing the two.

Emphasis on confidence intervals rather than P values

Confidence intervals are fairly easy to understand, while the convoluted logic of P values make them hard for many to understand. IB explains confidence intervals in detail (Chapters 4, 5, 6, 12 and 13) before even mentioning P values (introduced in Chapter 15). Every example emphasizes confidence intervals. BF and SLS also present confidence intervals before P values. PB and BA present P values first. MS presents them simultaneously.

What is a P value?

IB defines and discusses P values in chapter 15, and then discusses how to interpret low and high P values in Chapters 18 and 19. PB also explores the meaning of a P value. Surprisingly, the other books only define P values briefly, without a detailed explanation. 

What does it mean when result is statistically significant?

The conclusion that a result is statistically significant is often misunderstood. Chapter 18 of IB clearly explains what this conclusion means and doesn’t mean, using a Bayesian framework to take into account the context of the experiment. PB also briefly makes this point. The other books do not.

What does it mean when a result is not statistically significant?

The conclusion that a result is not statistically significant is also easy to misinterpret.  Chapter 19 of IB explains. PB also devotes a chapter to explaining what ‘not statistically significant’ means and doesn’t mean. The other book do not.

Multiple comparisons

The problem of multiple comparisons is one of the most difficult issues in statistics, and IB discusses the problem generally (Chapter 22 and 23), in the context of ANOVA (Chapter 40), and in the context of multiple regression (Chapter 38). BA, PB and SLS discuss multiple comparisons only as follow-up to ANOVA.  BS and MS barely mention the problem of multiple comparisons. None of the books other than IB mention the False Discovery Rate.

Survival data

The end point of many clinical trials is survival time. IB explains how to interpret survival data, and how censored observations are handled (Chapters 5 and 29). BF, MS and PB also explain survival data, but this topic is omitted from SL and BA.

Nonlinear regression

Many fields of biology (especially pharmacology, biochemistry, physiology) commonly analyze data by nonlinear regression. While IB devotes an entire chapter (36) to this topic, BA has two pages on nonlinear regression and the other four books don’t mention it.

Outliers

The presence of one or a few outliers can ruin a statistical analysis. Handling outliers is a difficult problem in data analysis, and IB devotes a chapter (25) to it. BA and MS briefly mention outliers. BF, PB and SL don’t mention the problem.

Testing for equivalence

Sometimes your goal is not to ask whether two groups are different, but rather to ask if they are equivalent. Standard statistical tests don’t approach this question in a useful way. Chapter 21 of IB explains how to answer this question properly. The other books don’t discuss testing for equivalence.

Lognormal distributions and geometric mean

Lognormal distributions are common. Chapter 11 of IB explains how these distributions arise and how to analyze lognormal data. IB also points out that high values are common in lognormal distribution and can be easily mistaken as outliers. BF does not mention lognormal distributions. The other four books mention it only briefly.

Normality testing

Chapter 24 of IB explains the use and limitations of using normality tests to determine whether data were sampled from a Gaussian distribution. SL and BS also discuss this topic, but the other books do not.

Meta-analysis

Chapter 43 of IB (new to the 3rd edition) explains the concepts of meta-analysis, a method used to combine the results of multiple studies. None of the other books mention meta-analysis.

 Reproducibility

IB has a new chapter on reproducibility. The other texts don't mention this important topic,

 Prices are in US dollars, as posted on amazon.com on Dec. 16, 2013.