Experimental design is a written strategy that describes clearly the specifics that are to be used when carrying out an experiment.The purpose of carrying out an experiment is to collect data; there is no prior knowledge of results of an experiment before it is conducted. Questions intended to be answered must be defined clearly before the experiment is carried out. Identification of anticipated source of variability in experimental specimen is vital because one of the objectives of experimental designs is to minimize impacts of the source of variability on design results.
In most of the experiments, certain factors are held constant and varying the other variable. However, this approach is inefficient when compared to altering the variable levels simultaneously. Key concepts in experimental design are controlling, randomization, and replication. When the cost of randomizing a variable is too high, controlling enables one to limit randomization by conducting trials with a particular setting of the variable. Randomization is the steps followed when conducting experiment trials; randomized sequence eliminates the effect of uncontrolled variables (Burman et al, 2010).
Consider a fictitious company that wants to carry out an evaluation of a new type of fertilizer before introducing it into the market. They apply the new fertilizer to a particular field (X), and at the same time apply a different type of fertilizer on field (Y). Field (X) receives adequate rainfall throughout the year and its mineral composition is well balanced whereas field (Y) does not receive as much rainfall as (X) and its soil is slightly acidic. During harvest season, field (X) yields more produce than field (Y). The conclusion made by the company is that their new fertilizer is more superior.
Effects of variations of the two fields were not factored when conducting the trials, this resulted in experimental bias. It cannot be deduced that the new product was responsible for the improved yield without controlling the variables. Randomization is used because it is almost impossible for the persons conducting the experiment to eliminate bias by only applying their judgment (Montgomery, 2013). Replication greatly improves the validity of randomized experimental results. Repetition reduces result variability with in turn increases the significance of experimental results.
Burman, L. E., Reed, R. W., & Alm., J. (2010). “A call for replication studies”. Public Finance Review, 787-793. doi:10.1177/1091142110385210
Montgomery, D. (2013.). Design and Analysis of Experiments. (8th ed. ed.). Hoboken, NJ:: John Wiley & Sons, Inc.