Customer Reviews

Common Errors in Statistics (and How to Avoid Them)

5 star:		(4)
4 star:		(1)
3 star:		(1)
2 star:		(1)
1 star:		(1)

 

 

Good:

-This text is written in a friendly, accessible style.
-Issues covered are relevant and solutions offered are practical.
-Frequent reference is made to the technical literature in support of arguments presented.
-Clear rules are articulated as to when one should consider using certain techniques.
-Material is fairly timely, including coverage of many recent advances in statistics.


Bad:

-By its nature, this book will involve the authors' opinions, with which we may not all agree. On the other hand, this is at least a good place to start the discussion.
-I wish the book were longer, giving the authors' the opportunity to cover more topics.

It is very common, and even necessary, to make mistakes while performing modeling or statistical studies. Some of these are easily corrected when one has the knowledge or expertise to understand that a mistake has been made. These could be designated as "blunders" and are part of the everyday life of a statistician or modeler. Other mistakes though are more serious, in that the investigator fails to recognize them and believes that the techniques used are valid for the problem that is studied. In addition, one can frequently fall into the trap of believing that mathematical or computational techniques or algorithms, whether done by hand or with the use of software, are always reliable and therefore require no independent checks or scrutiny. When statistical studies or modeling is done in an area where there is no danger to human or animal life, errors only have the effect of diminishing the validity of the study (and possibly the career of the investigator). In areas such as medicine and civil engineering however, errors in statistical studies can have serious ramifications for human life and safety, and therefore it is crucial that investigators be aware of how they arise and how to avoid them.

This short book is very valuable in that it discusses many of the errors that can arise in statistical modeling and gives advice on how to avoid them in practice. It should be on the bookshelf of all practitioners, regardless of their accumulated years of experience or level of expertise. Sometimes it is difficult for modelers to admit that they have made mistakes, let alone admit that they need advice for performing tasks they may been doing for years. But it never hurts to acknowledge that certain practices, even if they are carved in stone, may not be applicable to certain situations, and this book gives examples of this that are drawn from real world experiences. The authors are careful not to patronize the reader, but they do not hesitate to point out some of the misadventures that have occurred in statistical modeling. Throughout the book they caution against a religious attitude about computer software and mathematical formalism, and give explicit examples of how unquestioned use of these can result in serious errors. This goes hand in hand with their belief, usually only implicitly expressed in the book, that time constraints (such as emergencies) and deadlines may restrict an optimal statistical analysis from being conducted, but that any analysis done using improper tools should not be christened as such by the statistician community. The length of the book of course prohibits an exhaustive analysis of statistical studies that have gone awry but the authors include references for the curious reader.

A fast well written user guide for a technical profesional. Would buy again and recomend for most Quality Engineers in all fields.

Phillip Good and James Hardin often succeed in their endeavor to make their content accessible to an audience beyond that of "hardcore" statisticians. Despite their many applications in any modern society, statistics look unappealing to most people. Sometimes, both authors get lost in esoteric debates about some statistical topics that are of limited interest to a wider audience. Furthermore, Good and Hardin give too many examples that are related to the medical field. The authors could diversify their examples in a fourth edition of their treatise to further expand their readership. To their credit, Good and Hardin repeatedly warn their audience against the servile reliance on statistical software. Software users have to check the default settings to see if they are applicable to the application at hand. The authors correctly note that the most common error in statistics is to assume that statistical procedures can take the place of sustained effort. For this reason, Good and Hardin urge their readers not to let statistics and by extension statistical software do their thinking for them. In conclusion, "Common Errors in Statistics (and How to Avoid Them)" is a nice addition to anyone's modeling / statistical library.

People who analyze and work with statistics will get the most out of this book. The interested layperson will get plenty of useful insights from the examples, case studies and the outlines for best practices.

This book is chock full of hands-on advice, contains many examples of real life applications and discusses all the major aspects of working with statistics - from data collection to drawing meaningful conclusions. Without being too heavy on the math, it coveres data quality assessment, estimation, hypothesis testing and various statistical procedures and regression models.

It is more than an excellent reference book - it contains ample food for thought and guidelines for the proper use of statistical methods and the interpretation of statistical information.

I know Phil Good very well and I have taken online courses from James Hardin. These are very experienced statisticians who have done much research and consulting. This is the thrid edition of a very popular book. I think its popularity stems from its wisdom, experience of the authors, their writing style and the practicality and suscinctness of their expositions. It is unusual because it teaches statistics by starting out with common mistakes and misconceptions.

I think it improves with every edition. Some of my work is cited. For example starting in the second edition Phil Good being impressed with my bootstrap short course lectures decided to incorporation my list of common myths or misconceptions about the bootstrap. This is the only source that has it. It is not even in my bootstrap book. Also my work on the sawtoothed nature of the power function for exact binomial tests is included in this edition. I don't want to emphasize my contributions because they are just two examples amongst a wealth of material that follow the theme of the subtle nature of statistics that the novice need to appreciate.

Their is much coverage about biostatistics a specialty area of the authors. In this edition the authors have added a chapter on data quality assessment, a chapter on correlated data and new sections on factor analysis and Poisson and negative binomial regression. The latter two topics are specialty areas of James Hardin that he teaches with Joe Hilbe in an online course at [...]. Dr. Good also has an online site [...] where he teaches resampling methods and other topics.

I have not seen the first edition. I have received from Amazon the 2nd edition. The authors are salient to errors in statistics but not the material in their own book. For example, Good 2005 is cited on p42 but does not appear in the References. Similarly Wald 1980 on p82. On p144 Hardin and Hilbe 2002 is cited but the only reference is Hardin and Hilbe 2003. On the other hand the authors have left many older references (eg on surveys, p43) stand without noting the new helpful material.

Instead I highly recommend Abelson (1995): Statistics as Principled Argument.

There is a real need for a book on "common errors in statistics," but this one does not do the job. It reads more like a mediocre statistics textbook.