Reliability: Modeling, Prediction, and Optimization

CHAPTER 10

Design of Experiments and Analysis of Variance

10.1 INTRODUCTION

In Chapter 3, types and sources of data were discussed and basic methods of summarization and description were given. Sources of data include historical data, test data, vendor data, and so forth. In subsequent chapters, we looked at basic statistical inference regarding the parameter(s) of specified probability distributions: estimation (method of moments and MLE), confidence and tolerance intervals, and hypothesis testing. We looked at continuous and discrete distributions, complete and incomplete data, and grouped data.

In this chapter, we are concerned with the collection of test data, specifically in the context of structured experiments. The basic objective will be comparisons of two or more populations. The emphasis will be on comparison of population means, although comparisons of other characteristics (e.g., variances or fractiles) will be considered to some extent as well. For example, an adhesive may be bought from several suppliers and we may wish to compare average bond strengths of the different products. Problems of this type are also encountered in analysis of dynamic reliability. For example, we may wish to compare MTTFs for several product designs under consideration.

By a structured experiment is meant a plan for data collection under which the experimenter has control over the important experimental conditions and other factors that could affect the results. Design of experiments (DOE) ...