The Analysis of Random Effects in Modeling Studies.

TLDR: In this article, the concept of a random factor is discussed along with its application to and relevance for modeling research, and a practical solution to the problem is proposed, where the experimenter conceptualizes a study as providing the basis for a generalization to a larger population of models.

Abstract: SCHEIRER, C. JAMES, and GELLER, SANFORD E. The Analysis of Random Effects in Modeling Studies. CHILD DEVELOPMENT, 1979, 50, 752-757. Much of the research on modeling has used single models or, when multiple models have been used, a questionable data analysis has been applied. The analytic and conceptual difficulty revolves around the decision whether to treat models as a fixed or as a random factor in the analysis. We argue that in almost all cases the experimenter conceptualizes a study as providing the basis for a generalization to a larger population of models. Since this is the case, models must be analyzed as a random factor or a positive bias is introduced into the results. In this paper the concept of a random factor is discussed along with its application to and relevance for modeling research. Worked examples are provided, and a practical solution to the problem is proposed.