Stat Labs: Mathematical Statistics Through Applications
No TL;DR found
Abstract
in the penultimate chapter: identify goals, cause-effect analysis, modeling, costbene t analysis, and recommendations to management. Most chapters end with a checklist like this, and most of those lists (the calculus one being an exception) have a similar character of moving from goals and de nitions, to data collection, to data analysis, to validation, and nally to predictions or business decisions. This is the aspect of the book that I think is quite laudable and sound. It reects the author’sintelligenceandhowhe has usedhis education(PhDin mathematics) and other skills to advance a successful career (adjunct professorships, mostly in computer science, I believe; and work in several companies and government agencies such as the Navy and the CIA). There are no exercises, in case you wondered. The index is useless: numerous entries refer to page numbers greater than the last page number. Marilyn vos Savantwas claimed to be on page 143,but she’s on page 131.Evidently,a chapter somewhere in the middle was removed but the index was not reconstructed. The calculus advice, and other aspects that I view as bizarre, come from the fact that, while May probably took a statistics course or two in college, he is selftrained in implementing the methods. He has gured out all this stuff for himself, based on the tools and environment available to him. Based on what is said on pages 162–163, the tools consist of pocket calculators, spreadsheets, MathCAD and other symbolic mathematics software, the book Numerical Recipes in C {no citation is provided, but it is Press, Teukolsky, Vetterling, and Flannery (1993)}, and several specialized computer programs for neural networks and data mining. The author’s calculus advice is based on the fact that his calculator can do simple linear and polynomial regression. For multiple regression or nonlinear regression, his method is to write spreadsheet macros that compute the tted values and residual sum of squares (pp. 89, 92, 93); then use Excel’s Solver addin to minimize the latter. He seems to be unaware of another add-in distributed with Excel, the Analysis Toolpak, that can do multiple regression directly. That add-in is a poor tool indeed, and some may wish that more people were unaware of it; but the point is that the author is clueless about the huge array of statistical software that has been developed over the past 40 or so years. And in 12 or more years on the faculty at Virginia Tech, he seems to never have had any contact with their excellent statistics department. That has not kept him from writing a book on data analysis. The point I want to make, and the reason this book is important to those of you who teach, is that it is a case study in how statistics is often practiced in business, industry, and government. There are many people out there who are quite intelligent, and are doing generally good jobs of analyzing data—but have a very distortedview ofwhat is useful or even easy. And they have taken statistics courses from us! While May is of an older generation than our current students, little has changed in the teaching of most statistics courses except perhaps for the incorporation of spreadsheet methods such as those described above into business statistics courses. So for your edi cation, I recommend that you obtain this book and read it carefully; but keep it away from your students. And please teach those students how to use appropriate tools.