Design of Experiments as a subfield of statistics has its origins trying to understand systems and making decisions. In the context of food supply chains where agricultural and farm systems, packaging and processing plants, cold chains and food logistics, product design, warehousing, retailing, among many other steps and actors, come together to produce safe, healthy, nutritious, reliable, affordable, and sustainable food products. At each stage, many problems have to be solved and decisions have to be made from agricultural practices, scheduling, land use, costing, and pricing, at the farm and orchard level. A general outline is presented on the general workflow; illustrated through examples and a current review of some of the applications of design of experiments in the context of food production help provide some directives and guidelines. Global trends and challenges that influence both the food industry and the practice of design of experiments are discussed.
TopIntroduction
Design of Experiments (DOE) as a subfield of statistics has its origins trying to understand systems and making decisions based off of the results that are obtained after experiment execution. It has a clear process route: Problem definition, designing the experiment, running the experiment and collecting data, processing data, analyzing, and either continue experimenting or reaching conclusions. Yet, at each stage decisions have to made, some objectively others heuristically. In the context of food supply chains where agricultural and farm systems, packaging and processing plants, cold chains and food logistics, product design, warehousing, retailing, among many other steps and actors, come together to produce safe, healthy, nutritious, reliable, affordable and sustainable food products. At each stage many problems have to be solved and decisions have to be made from agricultural practices, scheduling, land use, costing and pricing, at the farm and orchard level; through the processing of fresh fruit such as cleaning, sorting, inspecting, and through the transformation of processed food, such as concentrating, pasteurizing, stabilizing, mixing, cooking; each step having problems and opportunities to maintain or exceed the market and customer demand.
Yet, the Design of Experiments is not always used to this end. This chapter will provide a practical overview of the Design of Experiments work flow. Provide some basic heuristics for each stage of the work flow; and illustrated in the context of the many stages of the food supply chain and its production systems.
At the problem definition stage issues related to identifying the problems and opportunities for improvement will be described. Heuristics on how to identify a problem that can be framed through the use of Design of Experiments will be provided. It is illustrated through case studies and examples in agriculture, food processing and food logistics. Common pitfalls will be described. Helping the Experimenter avoid costly mistakes, such as providing the basic insight to variable definition where for example, an experimenter takes 3 factors that influence an output, yet two are not of interest in the decision model, when the tools of blocking could provide a means to manage this arrangement.
For the designing of experiment stage a general decision algorithm will be presented in order to help define the type of problem in relation to the possible experimental design one can use. Knowing what techniques one can use, understanding the economics in terms of time, money and resources, and the reliability and validity of a design strategy will all be detailed and illustrated in this section. These issues are important in any DoE context, but hold special importance if food production given that products are meant for human and animal consumption, and because in many developing and underdeveloped countries small profit margins and unfavorable environmental and social conditions may create a restrictive environment for the application of experiments.
Basic considerations during the running of experiments and data collection will be described and illustrated. Issues on randomization, sample size, replication, measurement, experimentation scheduling, experimental execution logistics issues, data collection pitfalls, data table design, and other practical issues will be developed. Some examples of best practices in food production context will be explained. An example that could be developed is the measurement requirements, where for example chicken feed could be measured in a volume cup or weighed, and what reasoning could be used to choose on over the other. Some information on data sharing and data maintenance will be briefly reviewed. This will be done with an emphasis of knowledge management for reproduction and replication in future studies. Some useful information on open source and proprietary software will be discussed where cloud based services can be helpful.
Strategies for analyzing results will be developed using examples and illustrations provided in the previous sections. A generalized description of the hypothesis testing results during the development of the experiment will be clarified. An overview of conventional results presentation and how to generally interpret or use information resulting from a DoE calculation will be described. And a decision route on what actions to take depending on the results will be expanded on.