Customer Reviews

Data Reduction and Error Analysis for the Physical Sciences

5 star:		(7)
4 star:		(1)
3 star:		(5)
2 star:		(1)
1 star:		(0)

 

 

Robinson's second edition continues the late Bevington's tradition of clear and concise writing, making this book a priceless reference for scientists. Robinson has added discussions of modern problems such as resolving closely-spaced peaks in a spectrum. The new version also adds chapters on Monte Carlo techniques and maximum-likelihood analysis, both powerful tools for data analysis made possible by better computers.

The chapter structure has been modified considerably, so those who have grown comfortable with the first edition over the past decades may not be able to find things as easily. Other than that, most of the weaknesses are computer-related. Much has changed even since 1992.

Robinson added an appendix on graphical presentation. This sounds promising but is a pretty trivial discussion of when to use linear or logarithmic axes and the advantages of a historgram. Might be useful for a very young student, but these days playing with such things is easy in any graphing program.

Many of the computer code snippets have been removed. Most of them were only a few lines of code with lots of comment lines anyway. The codes that remain have been moved from the main text to a densely-packed appendix, which makes them more difficult to study while reading the text.

The codes themselves have been updated from old FORTRAN to a structured language, but I would have preferred C or FORTRAN 90 over the chosen PASCAL. The latter may be useful for undergraduate students, but I've never seen a PASCAL compiler in a working physics lab.

The included disk is a now-obsolete 5.25" floppy. I had to hunt for a machine that could read it and copy over to a 3.5" disc. The text claims repeatedly that the disc has both FORTRAN 77 and PASCAL routines on it, but my copy only has the PASCAL.

In the end, it's the textual content that is important, and this book is a fantastic basic discussion of data analysis and statistics for students and a great reference for the practicing scientist.

I make measurements frequently and this book is great for providing the background to analyze your data.

I took undergraduate level statistics and it never really gave the practical applied background in how to analyze data. It merely presented concepts and presumed you knew how and why to apply them. This book is very good at helping you to understand the how and why.

I have read a number of other statistics book in search of the practical applied information provided in this book and did not find it in the other books.

The writing is clear and consice. There is enough background provided for even those unexposed to statistics.

I have not tried the software. Most of the formulas are easy to apply and can be implemented in simple programs or spreadsheets in very little time.

In short, I recommend this book to anyone making measurements of any kind.

This book seems to have been completely rewritten by the new author, only keeping the outline of the original, and it's for the better. The writing is as careful as the original, and as economical, so you have to master the early chapters or the rest is hopeless, as things start off slowly but quickly become difficult. It begins by considering the error in a single measurement, and proceeds to estimating errors derived from curve fitting. A few nuclear decay experiments provide examples throughout, and the author insists on calculating many quantities manually, even though in practice it would never be done that way. Some background topics like matrix algebra appear in the appendix too.

Bevington's book "Data Reduction and Error Analysis" is an old war-horse that is now in its third edition, and looks a good intermediate-level, practical reference to have on the shelf. I bought my copy only recently to follow up a reference, so am less familiar with it than I'd like to be before writing a review.

Unfortunately, following that reference led me to equation 4.22, which is wrong. When forming a weighted average of variances, one needs to square the weights in the numerator and square the sum of weights in the denominator. The equation as given has the first power of the weights and of the sum, just as found in the weighted mean. As I sort the matter out myself (with help from a friendly statistician) and look around at other treatments, this appears to be a common mistake, but mistake it is.

Otherwise, it seems a good book clearly written.

As a practicing engineer, this was one of 5 or 6 books that I found most essential throughout my career. Chapter 4 on Propagation of Errors was referred to very often in my work.

Simply put, not as clear as John Taylor's error analysis book. If you want a good explanation for how and why to approach error analysis, this book does not do the job.

There are lots of misleading in Ch. 3. - Specific error formulas. See the original Bevington's 1st ed., and compare the formulas. I think this misleading should be corrected. Standard deviation is ALWAYS positive!!..

Bevington's first edition of this book dates to the late 60's when Fortran ruled the world. I was crushed when I lost my copy in the mid 70's and am delighted to find he's written a modern updated edition!

This is a book that has been widely used by three generations of students and researchers. I used the 1969 edition as an undergraduate. The 2003 paperback edition would be an equally excellent publication except for the astounding number of editing errors in formulas and a few gross errors that seem to be outright mistakes. One particularly egregious error concerning multiplication of linear matrices appears in appendix B. I bought the text as a refresher and reference. Because I had a working knowledge of the subject matter, most of the errors were immediately evident. However, someone new to the subject would find many of them confusing.

In My opinion, McGraw Hill has done a terrible disservice to the authors and should be embarrassed at doing such a poor job of publishing an otherwise excellent book.

I first read this book almost forty years ago when I was a physics major taking my first physics lab course. This book introduced me to such concepts as variance, Chi-squared, and least squares linear regression. This book was sufficient for my purposes until I was a graduate student and needed a more sophisticated estimate of the standard deviation for my experiments, which led me to Kendall and Stuart: Kendall's Advanced Theory of Statistics, Volume 1: Distribution Theory.