Designing Quantitative Experiments: Prediction Analysis

Early in my career I was given the task of designing a sub-critical nuclear reactor facility that was to be used to perform basic research in the area of reactor physics. We planned to run a series of experiments to determine fundamental parameters related to the distribution of neutrons in such s- tems. I felt that it was extremely important to understand how the design would impact upon the accuracy of our results and as a result of this - quirement I developed a design methodology that I subsequently called prediction analysis. After working with this method for several years and applying it to a variety of different experiments, I wrote a book on the subject. Not surprisingly, it was entitled Prediction Analysis and was p- lished by Van Nostrand in 1967. Since the book was published over 40 years ago science and technology have undergone massive changes due to the computer revolution. Not - ly has available computing power increased by many orders of magnitude, easily available and easy to use software has become almost ubiquitous. In the 1960's my emphasis was on the development of equations, tables and graphs to help researchers design experiments based upon some we- known mathematical models. When I reconsider this work in the light of today's world, the emphasis should shift towards applying current techn- ogy to facilitate the design process.

The method of Prediction Analysis is applicable for anyone interested in designing a quantitative experiment. The design phase of an experiment can be broken down into problem dependent design questions (like the type of equipment to use and the experimental setup) and generic questions (like the number of data points required, range of values for the independent variables and measurement accuracy). This book is directed towards the generic design phase of the process. The
methodology for this phase of the design process is problem independent and can be applied to experiments performed in most branches of science and technology. The purpose of the prediction analysis is to predict the accuracy of the results that one can expect from a proposed experiment. Prediction analyses can be performed using the REGRESS program which was developed by the author and can be obtained free-of-charge through the author's website. Many examples of prediction analyses are included in the book ranging from very simple experiments based upon a linear relationship between the dependent and independent variables to experiments in which the mathematical models are highly non-linear.