Data Analysis: Statistical and Computational Methods for Scientists and Engineers (Ohlin Lectures; 7) - Hardcover

Brandt, Siegmund

Hardcover

ISBN 10: 0387984984 ISBN 13: 9780387984988

Publisher: Springer, 1998

Synopsis

1. 1 Typical Problems of Data Analysis Every branch of experimental science, after passing through an early stage of qualitative description, concerns itself with quantitative studies of the phe­ nomena of interest, i. e. , measurements. In addition to designing and carrying out the experiment, an importal1t task is the accurate evaluation and complete exploitation of the data obtained. Let us list a few typical problems. 1. A study is made of the weight of laboratory animals under the influence of various drugs. After the application of drug A to 25 animals, an average increase of 5 % is observed. Drug B, used on 10 animals, yields a 3 % increase. Is drug A more effective? The averages 5 % and 3 % give practically no answer to this question, since the lower value may have been caused by a single animal that lost weight for some unrelated reason. One must therefore study the distribution of individual weights and their spread around the average value. Moreover, one has to decide whether the number of test animals used will enable one to differentiate with a certain accuracy between the effects of the two drugs. 2. In experiments on crystal growth it is essential to maintain exactly the ratios of the different components. From a total of 500 crystals, a sample of 20 is selected and analyzed.

About this title

Review

From the reviews:

"The book is concise, but gives a sufficiently rigorous mathematical treatment of practical statistical methods for data analysis¿It can be of great use to all who are involved with data analysis." Physicalia

"... Serves as a nice reference guide for any scientist interested in the fundamentals of data analysis on the computer." The American Statistician

Synopsis

This book bridges the gap between statistical theory and physcal experiment. It provides a thorough introduction to the statistical methods used in the experimental physical sciences and to the numerical methods used to implement them. An accompanying CD-ROM provides detailed code (in Fortran77 and in C) for implementing many of these algorithms (for IBM PS/2 or PC machines). The treatment emphasizes concise but rigorous mathematics but always retains its focus on applications. The reader is presumed to have a sound basic knowledge of differential and integral calulus and some knowledge of vectors and matrices (an appendix develops the vector and matrix methods used and provides a collection of related computer routines). After an introduction of probability, random variables, computer generation of random numbers (Monte Carlo methods) and impotrtant distributions (such as the biomial, Poisson, and normal distributions), the book turns to a discussion of statistical samples, the maximum likelihood method, and the testing of statistical hypotheses.

The discussion concludes with the discussion of several important stistical methods: least squares, analysis of variance, polynomial regression, and analysis of tiem series. Appendices provide the necessary methods of matrix algebra, combinatorics, and many sets of useful algorithms and formulae. The book is intended for graduate students setting out on experimental research, but it should also provide a useful reference and programming guide for experienced experimenters. A large number of problems (many with hints or solutions) serve to help the reader test

Edition details

• Publisher: Springer
• Publication date: 1998
• ISBN 10: 0387984984
• ISBN 13: 9780387984988
• Binding: Hardcover
• Language: English
• Edition number: 3
• Number of pages: 686