Randomization Statistic -- Research Methodology

Randomization

Statistic -- Research methodology

The importance of randomization in experimental designs

The importance of randomization in experimental designs

When conducting an experiment, it is essential to minimize the influence of extraneous and irrelevant variables upon the experiment's outcome. For example, when studying the effects of a new drug, if all of the subjects were male, of a similar age, and similar socioeconomic status, variables that pertained to their demographic group might appear as side effects or positive results of the drug, even though these effects were not directly caused by the treatment. It is impossible to anticipate every single external influence upon the participants of a study. However, through the use of randomized samplings, the tendency for one group's characteristics to influence the results can be minimized.

When an experiment uses a control group as well as an experimental group, the two populations should ideally be demographically balanced. If one population was much more affluent than the other population, it might appear as if the treatment or variable being studied had a powerful effect, even though the real source was due to external influences, such as the benefit of their higher income levels. Depending on the resources of the experiment or the size of the population in question, randomization may be achieved either through a genuine, completely random, computer-generated allocation of subjects, or through more careful dispersing of subjects into different groups to ensure a demographic balance. Complete randomization is most effective for very large sample sizes.

If an experiment's test subjects are small, or manifest significant characteristics amongst certain subset populations that could influence test results, the experimenter may use a randomized block design. "In a block design, experimental subjects are first divided into homogeneous blocks before they are randomly assigned to a treatment group. if, for instance, an experimenter had reason to believe that age might be a significant factor in the effect of a given medication, he might choose to first divide the experimental subjects into age groups, such as under 30 years old, 30-60 years old and over 60 years old. Then, within each age level, individuals would be assigned to treatment groups using a completely randomized design. In a block design, both control and randomization are considered" (Experimental design, 2010, Yale).

In stratified randomization, certain factors are declared to be potentially critical in influencing results and are allocated between the two different groups to ensure that only the experimental variable is emphasized (Johnson 2005). For example, when studying the effects of having fast food restaurants near schools and their effects upon student BMI, it would be easy to conclude that poorer neighborhoods tend to have more McDonald's, but that other factors besides the proximity of fast food affect poorer children's higher BMI (such as fewer supermarkets and fewer places to exercise). However, by stratifying the different groups, and having each group have the same number of children from various socioeconomic groups, the results would be more relevant if both poorer and wealthier children had higher BMIs if their schools were located near fast food establishments. This would suggest that proximity can encourage students to patronize unhealthy fast food restaurants in greater numbers, or normalize fast food consumption in their minds when they are outside of the school environment.