Factor, Fixed

Determining Whether a Factor Is Fixed or Random

Whether to consider a factor fixed or random can be a point of confusion in experimental design. One way to think about a fixed factor is that the selected levels of the variable are the exact ones of interest rather than those levels serving as a sample of all potential levels of interest for the variable. For example, a researcher may be interested in studying the effectiveness of five specific health campaign television commercials. There are many other potential commercials to choose from, but the effects of any other existing commercials are not measured or analyzed. If instead the researcher was interested in all possible health campaign commercials and had selected five random commercials as a sample intended to be representative of all existing commercials, the commercial would become a random factor.

Another way to determine whether a factor should be fixed or random is to imagine what would happen if the study were repeated. If the same elements were selected again, the factor is fixed. In the example of the health campaign commercials, if a repetition of the study intended to measure again the effect of those exact five commercials, commercial would be a fixed factor. If a repeated study used a new sample of five random health campaign commercials, then commercial would not be a fixed factor. Sometimes a fixed factor is defined as one in which effects are constant because its levels remain (or would remain) constant across replications of the experiment, rather than varying as samples of the population. In the example of the social media apps, if a replication study used Facebook, Twitter, and Instagram as the treatment conditions, this would be considered a fixed factor across studies.

...