Most helpful critical review
Great content poorly presented
on 28 March 2010
I approached this book because I needed to write a Fortran program for estimating a particular logistic regression model. I know Fortran 77 fairly well but I decided that it was time to switch to 95. Imagine my joy when I saw that the book explains Fortran 95 using the estimation of a logistic regression as the main example. However, I found it quite difficult to understand the nuts and bolts of Fortran 95 because of the many new concepts and definitions thrown at me before seeing a real example.
The book content is very relevant for a computational statistician: it takes the reader from not knowing Fortran 95 to writing a very complete and robust program for estimating a logistic model. It includes an extremely good explanation of how to create dlls callable from R, Splus, and SAS. It even explains how to create state of the art COMs (I haven't read this chapter carefully yet). The theory and math for the examples used are given with references and each sections has a set of problems. The code presented is provided through the web site.
The book assumes a prior knowledge of Fortran 77 or a similar language. The concepts are explained fairly clearly. The approach to coding is very rigorous and takes care of memory management, error handling and style issues. However, the book introduces many concepts without giving "hello world" type examples. So, when the first examples (I mean something that you can compile and run) are given, one is presented with quite complex programs, with all sorts of error handling procedures, portable statements, complex data structures, hidden data, etc. which are very hard to understand without having digested the basic concepts first. I found myself sifting through the examples looking for the lines of code that I needed to understand the basic functioning of the program and its structure. I believe that I would have understood better if I had been given the bare program/module/subroutines structures first and walked through the refinements little by little.
I gave only 3 stars to this book mainly out of the frustration it created in me but the excellent content and the care that the authors have put in writing it would deserve 5 stars. Had the book had more introductory examples, it would have been the perfect guide for any statistician who wants to use Fortran 95 in his work.