ABSTRACT
Up until this point, all of the analyses have been concerned with independent variables that may be considered fixed factors. A different type of factor is called a random factor. A random factor is one in which you have sampled some levels of the factor but think there are many more (actually, virtually infinitely more) levels that exist to which you wish to generalize. In comparison, a fixed factor is one in which all the levels of the variables that exist in reality and to which you wish to generalize are included in the study. For example, if gender was a factor in a study with two levels (men and women), those are all that exist in reality and that you wish to generalize to, so it is a fixed factor. Most of the factors in the social sciences are fixed factors (Keppel, 1991). Researchers typically use random factors primarily to expand generalizability of the research results. Often the effect of the random factor per se is of little interest. When random factors are included in the design, the error term for each F ratio may change. MANOVA can handle both random and fixed factors by permitting explicit specification of each effect’s error term on the “DESIGN” subcommand.
