1  Terminology

📖 Status of the book

Hi there! This book is a work-in-progress. You may like to come back later when it’s closer to a complete state. If you would like to raise issues or leave feedback, please feel to do this:

• at GitHub issues,
• email me at emi.tanaka@anu.edu.au, or
• leave it anonymously in the form here.

The field of experimental design is large and is applied to many domains, e.g. biology, psychology, industry, finance, marketing and so on. While some specialised terminology exist, the following terminology are basic terms that are commonly understood across domains:

1.1 Treatment related

• Treatments: the whole set of experimental conditions in which the units are subjected to.
• Treatment structure: the division of treatment levels to meaningful groups, e.g. factorial structure means all combinations across the different treatment factors.
• Control in treatment refers to a standard approach used, e.g. chemotherapy is a standard way to treat patients with particular cancers, and
• Placebo means a treatment designed to

1.2 Unit related

• Experimental unit: the smallest unit in which the treatment can independently be applied. If the experimental unit is a human, then each unit is often referred to as a subject.
• Observational unit: the smallest unit in which the response is measured on.
• Block or cluster: the grouping of another unit such that the units within the same block or cluster is more alike in some aspect than units in different blocks or clusters.
• Unit structure: the relation between the units, e.g. pots nested in particular glasshouses, subjects segregated by sex, and so on.

1.3 Record related

Record is used here to mean any uncontrollable factor. These can include the response, nuisance factor and just any value that is observed.

• Observation implies the observed value of a single unit.
• Response is usually the observation of interest in an experiment.
• Dependent variable is another way to call response.
• Nuisance factor
• Lurking variable
• Trait

Records are captured at different time points. We can map the time point categorical relative to the experiment: historical (pre-experiment), pre-treatment, post-treatment, followup/future (assuming measured once), or even the frequency of it: daily, weekly, monthly etc.

set_rcrds(yield = measure(plot, unit = "t/h", when = "pre"))

1.4 Experiment related

• Experimental run: an experimental setting to be used for collecting an observation,
• Design: the allocation of treatment to units,
• Plan or layout: the actual assignment of treatment to units, and
• Experimental structure: the arrangements or constrains of experimental factors – the structure can be broadly categorised as unit and treatment structure.

The above terminology describe the experimental factors and its associations. More specifically, experimental factors are any variables that pertain to the experiment (physical or otherwise) and includes the treatments, experimental units, observational units and blocks/clusters. The associations of experimental factors include its nesting structure or say the assignment of treatment to units.

In addition to above, there are commonly understood terminology that refer to the properties of experimental designs. These include:

• balanced experiment: refers to when the replication of the treatments are equal,
• orthogonal design
• complete block design: every treatment appears exactly once in each block or cluster, otherwise it is referred to as incomplete block design,
• blinded experiments, relevant only to studies involving human subjects, refers to experiments where subjects are not informed about which treatment they received. Double-blinded studies means that both those administrating and receipient of the treatment are not informed about the actual treatment.

plant growth

• The experimental aim is to study the effects of plant growth of four particular bacterial strains.
• The experimental units are …

1.5 Specialised terminology

Experimental design is applicable across numerous domains, so perhaps it is not surprising that there are domain specific or specialised terminology that is used to describe the whole experimental designs, e.g. Latin Square Design, Step-Wedge Trial, Beehive Design and so on. Some recipe of these named experimental design are presented and discussed in ?sec-cookbook.