Assessing the Effect of Time in Factorial Designs: The Practical Application

摘要

The scientist asks a statistician the question: 'should we incorporate time as a factor in our factorial designs or measure the response at various time-points?' Advantages and disadvantages of the approaches are discussed. The aspects to be considered include: the relative costs of performing the experiment, the response measurement and the statistical analysis; the quality of the data collected and the information to be gained from the experimentation; implications for possible designs; implications for setting up, analysis and interpretation of designs in standard software available to scientists; and pitfalls to avoid. The field of chemical process investigation is used to illustrate some of the issues involved. In some scientific areas the experimental or resource issues may be overriding and the answer obvious. For the chemist, in some cases the only choice will be the factorial design which includes time as a factor due to experimental reasons but in most cases the appropriate answer depends on careful consideration of the aspects listed above. The paper illustrates the importance of a partnership between scientist and statistician in addressing the practical application of statistical techniques.