The Design of Experiment or DoE for short refers to that branch of applied statistics which deals with planning, conducting, analyzing and interpreting results of controlled tests with the objective of assessing the outcome in response to variables due to changes in one or more factors. The original work in this field was done by R. A. Fisher in the early 20th century. He argued that a properly designed experiment and execution leads to avoiding frequently encountered problems in analysis it later.
A true DoE is characterized by -
- Understanding the effect of change in any input factor on the outcome
- Develop a good fitting mathematical function that gives the relationship between input and output under carefully controlled conditions with a greater degree of confidence.
Experiments differ from a natural phenomenon in the fact that in the case of experimental work, researchers play an active role rather than being a passive observer.
A good clinical trial design ensures reduced bias and in-accuracy in predicting treatment effects irrespective of statistical analysis. When a trial is well-designed and properly executed, statistical analysis can be performed, modified or even corrected if necessary. The shortcomings of a poorly designed trial cannot be improved after the trial is complete.
Controlled clinical trial contains all the key elements of a true experimental design. When these trials are conducted with proper randomization, they instill confidence that bias has been minimized. They can be easily replicated.
A truly controlled experiment is characterized by.
Researcher varies (manipulates) one of the independent variable (experimental treatment or intervention) by administering treatment to some subjects and withholding it from others. The resultant effect on the dependent variable is observed and recorded.
Control group serves as a standard or base for ascertaining the effectiveness of any treatment. A comparison is made between the results obtained from the experimental group (group that received the treatment) and control group (group that did not receive the treatment) to nullify the effect of external factors (factors other than the drug)
Randomization involves assigning the individual subjects to different groups (experimental and control) on a random basis. Random means each subject has an equal chance of getting assigned to any group. There is no systemic bias during the assignment.
There are three main types of experiment designs.
- Basic Design
In a simple design, the dependent variable is measured only after the administration (post-test only). In more complex designs, the dependent variable is measured at two points, before (pre-test) and after (post-test) the experimental intervention. However, the researcher manipulates only one variable at a time.
- Factorial Design
In this type of design, the researcher manipulates two or more independent variables at a time. This allows for testing multiple hypotheses at the same time. It allows the researcher to know the main effects (from an manipulated variable) and interaction effects (resulting from interaction of independent variables). It is impossible to get these results by conducting two separate experiments manipulating one independent variable at a time.
- Cross-over or Repeated MeasuresDesign
In this design, the same subjects are exposed to more than one treatment in random order. Patients receive different treatments during the different time periods, i.e. patients cross over from one treatment to another during the course of the trial. This contrasts with a parallel design, in which patients are randomized to a treatment and remained on that treatment throughout the duration of the trial.
We can conclude by saying that a properly designed clinical trial will help in determining the efficacy and effectiveness of a treatment in real-world allowing for proper statistical analysis of the result.