Random Assignment

Necessitation of Randomness

As part of the basis for experimental research, random assignment is applied to the selection of participants for treatments in order to remove any possible influence from confounding variables (i.e., unknown variables that could influence the relationship between known variables being tested). This is possible because the variables in the groups have been controlled according to whatever the research design may be. For instance, the experiment from the prior example is interested in understanding how well people retain information (X) while listening to music via headphones (Y). This experiment is most likely carried out in the controlled environment of a laboratory where outside distractions can be reduced, and the only variable that should influence differences in the observations or scores of information retention (X) is listening to music via headphones (Y).

Random assignment begins to establish the necessary parameters for determining causation by highlighting a direct linear relationship between [Page 1398]variables (i.e., when variable X changes, a change in Y also occurs). Statistical testing, such as linear regression, can determine whether or not one variable is causing the other, but the experimental design must be properly laid out and include random assignment for any meaningful causation to be determined. On occasion the researcher can make a mistake in randomly assigning participants, and assignment bias completely negates any results or causal relationships found during analysis. The goal of random assignment is to create two groups in one sample that are homogenous except for the experience of the treatment. When assignment bias occurs, there are differences between groups other than the experience of an experimental treatment. The treatment cannot be singled out to represent the cause of these differences, so any further analysis on the data is useless. In addition to providing the necessary foundation for causation, random assignment also assists with the internal validity of a study. Internal validity refers to the idea that the research design is measuring what it purportedly set out to measure. For example, if a researcher proposes that fewer hours of sleep (Y) is a cause of quality of work (X), then controlling the conditions that the variables are tested in becomes paramount. Remember, random assignment is trying to establish linear relationships between the variables so that no outside interference (i.e., confounding variable) hinders the understanding of the cause-and-effect relationship. In this example, the experimental group would sleep fewer hours than normal and then perform some work-related tasks to be graded on a scale. Ensuring validity, the control group would sleep their normal amount of hours and then perform the same work-related tasks, so that the only main difference between the two sample groups is the experience of the treatment. When performed correctly, random assignment guarantees internal validity and supports the existence of causal relationships.

...