No abstract

Although the concept of moderator variables has been used extensively in marketing-related studies, much confusion persists as to how they are defined and identified. To alleviate this confusion, the authors present a typology of moderator variables with a framework for identifying their presence and type. Simulated data are used to illustrate and validate the proposed framework.

No abstract

Practical theoretic means for assessing the sampling variability of loadings estimated by exploratory factor analytic procedures have not been readily available in the absence of restrictive distributional assumptions. It has been necessary for researchers to interpret these point estimates (loadings) through the use of arbitrary rules-of-thumb. Under these conditions, loading interpretations may be problematic. A method is presented for exploiting information in the empirical data, collected for a study's primary goals, to approximate confidence intervals for factor loadings, The method appears generalizable across factor methods, numbers of extracted factors, and rotation criteria.

Multivariate multiple regression, canonical correlation, and redundancy analysis are compared in terms of their analytical capabilities in behavioral research. Redundancy analysis is emphasized in the comparison. Both verbal and mathematical descriptions of these procedures are presented. The discussion of these procedures is followed by two illustrative examples. The first demonstrates the dissimilarities in statistical results that may be produced by canonical correlation and redundancy analysis, as well as multivariate multiple regression. The second illustrates circumstances that can lead to similar canonical and redundancy results, an uncommon but possible occurrence in behavioral research. Then, criteria are listed to assist analysts in selecting among the three procedures.Behavioral research frequently involves examining relationships between two sets of variables. Such instances might include analyses of relationships between personality traits and aspects of behavior, of multifaceted constructs of mood and social judgments, and of measures of affect and attributions to self and referent others. Several alternative analysis procedures might be used when two sets of multiple variables are involved. Which procedure is most advantageous and appropriate depends on the conceptual framework that forms the basis of the investigation, the research questions being addressed, the scale properties of the data, and the capabilities and limitations of alternative methods, as well as other possible considerations.Three analysis procedures-multivariate multiple regression analysis, canonical correlation analysis, and redundancy analysis'-are compared in this article because the existence of certain commonalities may result in some uncertainty as to which method would be most appropriate for a particular analysis. A more technical comparison is provided by Muller (1981), which the reader may consult for derivations and more extensive mathematical statements. Redundancy analysis is emphasized because it is the newest of the three procedures (Fornell, 1979;Johansson, 1981; Muller, 1981; van den Woilenberg, 1977) and may be less familiar to some readers. The emphasis on redundancy analysis and the discussion of only these three algorithms should not be interpreted as implying that they, individually or as a group, are always superior in interset analyses.