Letters: May 2020

I read the article ‘Smart trial and error for rapid innovation’ (Chemistry World, March 2020, p5) with much interest and am surprised that chemists involved in experimental work are still reluctant to move on from changing one factor at a time.

In the 1960s I was encouraged by a mathematician and a statistician to use factorially designed experiments involving changing several factors at a time to account for interactions. The practical work immediately became much more interesting and gave more valuable results than those obtainable by the ‘change one factor at a time’ method. The mathematics is not difficult but seemingly still stops chemists changing their ways. At about the same time cumulative sum techniques were also introduced for detecting changes in the average level of a sequence of figures, determining the point of onset of such changes, obtaining a reliable estimate of the current average value and making short term predictions of the future average level. A most useful investigative and reliable tool.

The Royal Institute of Chemistry organised a summer school in the quantitative treatment of experimental data in chemistry in 1971, which covered the scope of statistics, various statistical tests, cumulative sum techniques and the design of experiments. This last topic included a factorial experiment with replication and three factors, each at two levels.

Nowadays, with the much greater use of computers the mathematical calculations would be even easier. Enter the results into a program and job done. After 60 years I find it difficult to believe that these statistically based methods are not in common use.