Design of experiments

The experimenter, whatever his field of study, is always faced with the difficult problem of how best to organize his trials. How to obtain the right information in the shortest possible time and at the lowest possible cost? That's the question we'll try to answer in this article.

Scientists have only been tackling this subject for a few years. The first to tackle the problem were agronomists and statisticians. The techniques and concepts they have developed are so general that they can be used in all fields. In particular, analytical chemistry offers a vast field of applications.

This science of test organization is a recent one, starting with the work of R.A. Fisher (early twentieth century). Strange as it may seem, it doesn't yet have a name. We've suggested Experimentalism or Experimentology, but the scientific community has yet to decide.

The aim of this new science is to optimize the choice of trials and their sequence during experimentation. We shall see that this goal can be achieved provided the experimenter follows a rigorous method and is willing to abandon certain habits. Once he has appreciated the power and validity of this new technique, he will become a fervent advocate of it.

We have personally found that the experimental design method is at least three to four times more efficient than the usual approach to conducting trials, i.e. it achieves the same results with three to four times fewer trials. What's more, this method provides experimenters with a powerful tool for reflection and analysis, enabling them to conduct their experiments with confidence and precision.

Experimental designs were first used in agronomy. Then, little by little, they were used in other technical fields. Chemists adapted them to their own problems. Recently, quality managers have discovered these techniques and are now making extensive use of them. The universality of these methods should lead to their use in many fields. We can already point to successes in fundamental research, applied research, industrial development and even manufacturing. Personally, we have advocated their use to reduce the number of computer runs in computer simulations.

The field of applications is extremely vast, and everyone's imagination is free to find new areas and new uses. In analytical chemistry, they can be used to find the optimum setting for an instrument, to discover the factors influencing the result of an analytical method, to improve circular assays, to detect systematic errors, and so on.