An experiment imposes an action treatment on a group of objects for observing the reaction. Because the validity of an experiment is directly affected by planning and execution, experimental design is extremely crucial for the validity of any experiment.
In experiments the researcher directs an action on to experimental units. These experimental units can be objects like any things or even subjects like humans or any organisms. For example, a population of certain illness is divided into few groups. Each group is 'treated' with a different dose of a medicine to checkout which dose is effective.
Biostatisticians study and analyse data generated from experiments. They have to plan their experiments properly to ensure that the right type of data is available to answer the questions. Moreover the data must be sufficient enough to be clear and efficient. This process of planning of the experiment is called experimental design.
Before carrying out the experiment, the researched must have a clear understanding of the data which he needs to collect and the end result which must be achieved. Also the researcher must be aware of the possible variations in the result. Thus, designing the experiment is very important in order to improve the precision of answers.
History of development of Designing of experiments
James Lind performed well-organized experiments in 1747 and he developed a cure for the disease scurvy. He was serving as a surgeon on a ship and he selected 12 men from the same ship. All these 12 men were suffering from scurvy. He then divided these men six groups of two each. For two weeks, he served various supplements to these people, along with their basic diet. These supplements were the treatments for scurvy as proposed by various people. He then compared the groups to find the efficacy.
The one thing missing out from his experiments was randomization. Randomization is very crucial in modern day experiments. Lind could get good and supportive data from his experimental design and he could find that one of the remedy worked very well.
Principles of experimental design
In 1935, Ronald A. Fisher, proposed a methodology for designing experiments, in his book “The Design of Experiments”. Design of experiments is basically the process of planning, designing and analysing the experiment in such a way to effectively draw valid conclusions. Powerful statistical methods are very important to get statistically sound results. The three important principles of experimental design are randomization, replication and blocking. These principles play a very crucial role in decreasing or removing experimental bias.
Randomization: A Randomized experimental design assigns the randomly to an experimental group. As it is very difficult to reduce bias just be expert judgment, randomization is a common practice. With the help of randomization most consistent and trustful treatment groups can be created. Sometimes it is difficult to completely randomize an experiment, due to lack of time and cost.
Replication: Replication is the process of repeatedly running a part of the experiment or a part o it under varying conditions. Repetition has three main properties.
- It allows the researcher to get more accurate estimate of the experimental error.
- It can give the more accurate estimate of the interaction effect
- It can help in decreasing the experimental error and increase precision.
Though randomization insures treatment of different experimental groups is as similar as possible, the results of a single experiment cannot be easily accepted. Randomly picking three objects from a group of five and experimenting on them may not be effective. To increase the effectiveness repetition of an experiment on a large group of objects is needed.
Blocking: Elimination of all the possible unnecessary deviations is very critical for any experiment. Arrangement of experimental units into blocks of similar entities is called as blocking. Observations collected under similar experimental conditions are grouped in the same block. Blocking reduces sources of deviations and thereby bringing more precision to the experimental design.
Steps for designing an experiment
Analysis of the design of experiments was built on the foundation of the Analysis Of Variance (ANOVA). The following are the steps for designing an experiment. Of the seven steps given below, step 1 and 2 are pre-experimental planning.
- Recognition of the problem
- Selection of response variable
- Selection of factors, levels and ranges
- Selection of experiment design
- Performing the experiment
- Analysis of the data statistically
- Final conclusions and recommendations
Step 1: Recognition of the problem- Before designing the experiment it is very important to recognise and clearly define the problem based on evidences. A list prepared with specific problems to be dealt in the experiment can be of great help.
Step 2: Selection of response variable- A variable which can give useful information about the problem to be studied must be selected. Generally standard deviations and average deviations of measured characteristics are taken as response variable. These response variables should be measures before and after conducting the experiment.
Step 3: Selection of factors - The researcher should know about the factors which may influence the performance of the experiment. These factors which may have an effect on the experiment are called trouble factors or nuisance factors. These are the factors which the researcher does not want to deal with or wants to fluctuate in the experiment. There are also some other factors which the researcher wants to be constant which are called as design factors. The trouble factors are of three types namely controllable factors, uncontrollable factors and noise factors.
Step 4: Selection of experiment design- While selecting experiment design the researcher must consider inferences from statistics and sequential analysis. Based on these the experimental design must be randomized.
Step 5: Performing the experiment- When all the above steps are well planned, there is no doubt about the smooth flow of the experiment. Also the results and data obtained from the experiment must be properly recorded and analysed statistically.
Step 6: Statistical analysis of the data- To arrive at valid conclusions, the data recorded must be analysed statistically. Though non-statistical methods are also followed statistical methods are more reliable.
Step 7: Final conclusions and recommendations- Depending on the conclusions drawn from the statistical analysis of the experimental data further recommendations are made.
Applications of designing the experiment
- The main aim of the designing of the experiments was limited to agricultural needs. But the modern designing of the experiments has spread across various streams.
- Designing of experiments finds place in almost all the areas of science and engineering. Mainly the statistical analyses carried out are very helpful.
- In recent years, there has been a considerable utilization of designed experiments in many other Designing of experiments has also been used in service sectors areas like business, finance and government operations.
- Share with your friends! -