Mixed Level Design

A mixed level design is one that uses a combination of fixed and random factors. A fixed factor is one whose levels are chosen based on theoretical or practical interest, and the researcher wishes to draw conclusions about those particular conditions or levels. Examples include whether a message is one- or two-sided and whether arguments are strong or weak. A random factor is one whose levels are selected randomly from a larger population of options. The chosen levels are assumed to be representative of the population of interest, and the researcher is interested in statistical generalizations back to that population. The most common example is the participants factor in an experimental design. Participants are presumed to be randomly selected from a population of interest and results are generalized back to similar participants. In communication research, common random factors include specific messages or arguments that are selected from a larger population of options and represent examples of message or argument types. Another way of thinking about fixed versus random factors is whether the levels or conditions would be identical in a replication. If replicating a persuasion experiment, the researcher likely will retain the levels of message sidedness as one- versus two-sided and argument strength as strong versus weak but would collect data from a different sample of participants using different messages and arguments. Levels of fixed factors remain constant across studies whereas levels of random factors vary across studies. In practice, a mixed level design typically involves at least one fixed factor and one random factor that is not the participants factor. This entry identifies four types of mixed level designs, their advantages and disadvantages, and their implications for validity and analysis.