ESSCA: A Multidimensional Analysis Tool for Marketing Research George M. Zinkhan University of Houston William B. Locander University o f Tennessee Multidimensional analysis techniques such as ESSCA (External Single-set Components Analysis) are useful for marketing researchers who want to estimate the dimensionality of a group of related measurement instruments. Here, the advantages and disadvantages of this procedure are illustrated through an investigation of four advertising recall measures. The ESSCA solution suggests that two dimensions of recall are actually being measured:favorable recall of stimulus features and brand name recall. INTRODUCTION Examining the dimensionality of stimulus items (ads, products, concepts) is one important application of marketing research tools. However, our understanding of multivariate tools is critical when addressing various dimensionality problems. Exploratory factor analysis or principal component analyses are most commonly used in situations where there are no specific hypotheses about the structure of the data set. In cases where there is strong a priori knowledge, confirmatory factor analysis or structural covariance analysis would be more appropriate. When the analysis task falls somewhere in between exploratory and confirmatory analysis, then other procedures may be considered, such as canonical correlation, redundancy analysis, or External Single-Set Components Analysis (ESSCA). All three techniques 9 1988, Academy of Marketing Science Journal of the Academy of Marketing Science Spring, 1988, Vol. 16, No. 1,036-046 0092-0703/88 / 1601-0036 involve multiple criterion variables and multiple predictor variables. The purpose of this paper is to discuss marketing applications of the most sophisticated of these procedures: ESSCA. By way of contrast, all three multiple criterion/ multiple predictor (MCMP) techniques are discussed. Then, with this background, the use and limitations of ESSCA are illustrated through an example drawn from an advertising research problem. MULTIPLE CRITERION/MULTIPLE PREDICTOR TECHNIQUES Canonical correlation, redundancy analysis, and ESSCA represent advances over more traditional techniques such as multiple regression since the former are not merely simultaneous forms of analysis but are truly multidimen- sional in nature. Under the multiple criterion/multiple predictor techniques, theoretical constructs (or variates) are explicitly modeled; multiple indicators (or measures) are allowed for each construct. Thus, there is a more sophisticated interplay between data and theory when compared with more traditional analysis procedures. In addition, all three MCMP techniques provide statistics which can be used to test the overall fit of a model as well as the significance of individual parameter estimates. A brief description of each of the MCMP techniques follows. Canonical Correlation Analysis Given two sets of variables, canonical correlation forms variates which maximally correlate. This is shown graphically in Figure 1. Once this first set of variates is obtained, a second set of variates can be extracted- orthogonal to the first. In order to test hypotheses, metric data are assumed. Canonical analysis creates variates through a formative and additive process; and in this way assumes a linear and additive relationship between two sets of variables. The estimation procedure is non- JAMS 36 SPRING, t988