What is experimental design in statistics

Experimental design

Author: Hans Lohninger

When we analyze data, we can distinguish between two basic types of experiments: (1) Observers Experiments that only allow the experimenter to watch and listen without being able to influence the variables observed. (2) In contrast, one can designed Perform experiments that allow control over the level of the variables used in the experimental set-up. Although in many practical situations the experimenter has no way of controlling the variables, it is instructive to learn about experimental designs and how to analyze the data obtained.

Here is a collection of the most important terms that apply to experimental designs (sometimes called factorial experiments):

Response variableThe (dependent) variable of interest; it is determined by the result of the experiment.
independent variablesVariables of the experiment that determine the experimental setup; In the literature, independent variables are also called factors.
FactorsVariables of the experiment that determine the behavior of the experiment; Factors can be qualitative or quantitative.
Factor levelThe values ​​of a factor used in a specific experiment.
EditsIf more than one factor is used, each processing corresponds to a specific combination of factor levels of all factors involved.
experimental unitAn experimental unit is the object on which the response is observed.

If the experimenter controls the processing and the assignment of experimental units to the specific processing, one speaks of a designed experiment.

What is important in experimental design is the way in which the factor levels are defined and the experimental units assigned to the individual processes. Several possibilities are possible:

  • completely randomized design
  • random block design
  • factorial experiments

A major tool for statistical analysis of experimental designs is analysis of variance (ANOVA).