Statistical Analysis of Measurement Errors

The nature of Exxon Nuclear's business of fabricating uranium fuels for commercial power reactors requires replicate measurement of difficult-to-measure characteristics. Errors of measurement are of dominant importance in the company's quality assurance program and in implementing special nuclear materials safeguards. This Exxon monograph provides a comprehensive, user-oriented approach to the statistical analysis of replicate measurement data. In this book the author has developed a general model of statistical inference based on maximum likelihood for experimental situations in which a multiple number of items are each measured by several different measurement methods. While the method of maximum likelihood has been well known to have optimum statistical properties, its application to multiple item, multiple measurement situations has been limited by the unavailability of computation methods. The method of maximum likelihood is compared to other methods of estimation with the advantages and disadvantages of each pointed out. Experience with the use of the maximum likelihood methods is demonstrated with numerous examples.
Although the method of maximum likelihood in this problem context was developed for use in the nuclear field, the subject matter of this monograph is by no means limited to this field of application. Rather, it has broad application in manufacturing and research facilities, wherever measurements are performed, and is certainly of importance to the petroleum and chemical industries.