The Influence of Involvement on Brand Dissimilarities and Mds Solutions

Citation:

Tammo H.A. Bijmolt and Michel Wedel (1993) ,"The Influence of Involvement on Brand Dissimilarities and Mds Solutions", in E - European Advances in Consumer Research Volume 1, eds. W. Fred Van Raaij and Gary J. Bamossy, Provo, UT : Association for Consumer Research, Pages: 148-153.

European Advances in Consumer Research Volume 1, 1993      Pages 148-153

THE INFLUENCE OF INVOLVEMENT ON BRAND DISSIMILARITIES AND MDS SOLUTIONS

Tammo H.A. Bijmolt, University of Groningen, The Netherlands

Michel Wedel, University of Groningen, The Netherlands

[This research was sponsored by the Economic Research Foundation, which is part of the Netherlands Organization for Scientific Research (NWO).]

SUMMARY

Multidimensional Scaling (MDS) methods have often been used in consumer research. Despite this frequent use, little is known about factors influencing the judgment of dissimilarities by consumers and the solutions provided by MDS algorithms. In this paper we describe the results of an empirical study which investigates the effects of consumer involvement on perceived dissimilarities between brands and the perceptual maps obtained from them. In the study we measure both brand-decision involvement and product involvement along with brand dissimilarities. The effects of involvement are examined using probabilistic MDS.

1. INTRODUCTION

Multidimensional Scaling (MDS) methods have often been used in consumer research for perceptual mapping of competitive brands (Cooper 1983; Green 1975). The number as well as the complexity of the MDS models that can be used increases. Nevertheless, a marketing researcher may have several difficulties applying MDS and interpreting the results. A large number of factors may influence the dissimilarity judgments made by subjects and the solution provided by the MDS algorithms. In previous studies some evidence has been found that factors related to the responding subjects (Johnson, Lehmann, and Horne 1990; Malhotra 1990), the data collection procedure (Jain and Pinson 1976; Malhotra, Jain, and Pinson 1988), the type of data (Humphreys 1982; Whipple 1976), and the MDS algorithm (Spence 1972; Weeks and Bentler 1979; Whipple 1976) affect the MDS solutions obtained. Yet, the impact of many factors is unknown, and there remains a need for studies investigating when and how different data collection methods and MDS models can be applied. In this paper we study the influence of the consumer involvement on paired comparisons data and the solutions of probabilistic MDS models.

The concept of involvement has been seen as one of the most important factors influencing consumer behavior and therefore moderating the response of consumers to marketing instruments. The relations between consumer involvement and among others cognitive processing (Celsi and Olson 1988; Miniard et al. 1991), the effects of advertising (Petty, Cacioppo, and Schumann 1983; Kirmani 1990), and brand choice (Gensch and Javalgi 1987) have well been established. But little is known about the perceptual differences between consumers with different degrees of involvement. As far as we know, none of the studies described in the literature investigates whether these differences in searching and processing information caused by differences in consumer involvement actually result in differences in perceptions.

The elaboration likelihood model (ELM) of Petty and Cacioppo (1984 and 1986) distinguishes central processing of information, and peripheral processing of information. The ELM identifies a person's level of involvement as one of the factors influencing the way information about products and brands is processed. It is expected that a high involvement level causes a consumer to centrally process brand-relevant aspects (see also MacKenzie and Spreng 1992). Therefore, a highly involved consumer is expected to have a well-considered perception of the brands, based on a relatively large amount of important features of the brands (Gensch and Javalgi 1987). On the other hand, low-involved consumers are hypothised to have a perception that is less well-considered and based on less relevant characteristics. Furthermore, low involvement might reduce the ability and willingness of subjects to perform the relatively difficult paired comparisons task, and thereby influence indirectly the data and the solution of MDS. Many MDS models assume perceptions to be homogeneous across consumers, and aggregate the dissimilarity data over the entire sample of subjects. Ritchie (1974) noted that in many occasions there are important differences between the perceptions of consumers, but he did not succeed in explaining these differences by personal attributes of the consumers. The main question of this study is whether differences in perceptions can be explained by differences in the involvement level of consumers. We will describe the results of an empirical study in which we measured the perceived dissimilarity between brands, and the involvement of the consumers with respect to two products, namely shampoos and automobiles.

2. DEFINING AND MEASURING INVOLVEMENT

There are many definitions of involvement available in the consumer and marketing research literature. We follow Engel and Blackwell (1982), who defined involvement as "the action of extended problem-solving behavior when the act of purchase or consumption is seen by the decision maker as having high personal importance or relevance". Involvement has a number of different sources and effects (Bloch and Richins 1983). To measure the complex involvement construct, it will be necessary to take each of these forms, sources, and effects of involvement into account. Therefore, single-item measures cannot capture the entire involvement concept, and a multiple-item set has to be used (Laurent and Kapferer 1985). Multiple-item measures of involvement should be valid, stable, and applicable for different products and situations. Throughout the years, several authors proposed multiple-item measurements of the involvement construct. Kapferer and Laurent (1985, 1986) proposed to measure an involvement profile instead of a single involvement score. It consists of the following factors: the product's pleasure value, the product's symbolic value, risk importance, and risk probability. Thus, the involvement profile of Laurent and Kapferer is much more elaborated than a single-item involvement measure, but it does not cover all the causes and aspects of consumer involvement, as was argued by Mittal and Lee (1989). Zaichkowsky (1985) developed the Personal Involvement Inventory (PII) to measure the concept of product involvement. Although Zaichkowsky developed her PII for product involvement, she suggested that the scale is also applicable for brand-decision involvement. Mittal (1989), however, expressed serious doubts concerning the use of the PII as a brand-decision involvement scale, and noticed that no measurement scales for brand-decision involvement were available. Therefore, he developed and tested a four-item scale to measure this aspect of involvement. The scale Mittal proposed consisted of the following items: degree of caring, perceived brand differences, importance to right brand selection, and concern with the outcome. Like previously described scales, it measures only part of the many aspects of consumer involvement.

Recently, Mittal and Lee (1989) integrated the above-mentioned attempts to measure involvement. They proposed a theoretical framework which included two forms of involvement, namely product involvement, the interest a consumer finds in a product, as well as brand-decision involvement, the interest of a consumer in making the brand selection. In order for a consumer to be involved with a product, the product has to satisfy certain goals of the consumer. Mittal and Lee (1989) distinguished three sources of involvement, namely utilitarian, sign-value, and hedonic goals. Utilitarian goals can be satisfied by the physical performance of a product when a consumer uses that product. Sign-value is the symbolic value of a product to a consumer. By gaining, possessing or using a product with a high sign-value, the consumer may achieve social and self-concept related goals. Hedonic value corresponds to the pleasure and affect provided by a product. Mittal and Lee constructed a causal model in which the two forms and three sources of involvement are measured and linked with each other. Each form-source combination and also each form itself is measured with three items, resulting in a set of twenty-four items (listed in appendix A). We use this set of items in order to measure product involvement (12 items) and brand-decision involvement (12 items).

3. HYPOTHESES

As mentioned in the introduction, on the basis of the ELM of Petty and Cacioppo (1984 and 1986) it is expected that the differences in product and brand-decision involvement between consumers result in differences between these consumers in perceptions of product attributes. The effects of involvement can be twofold: effects on the perceptual map in the mind of a consumer, and effects on the mental operations of a consumer during the data generating process. In an empirical study, these effects are hard to separate. Instead of the distinction mentioned above, we distinguish between effects on the dissimilarity data themselves, and effects on the perceptual maps derived from them with MDS procedures. These effects are related, since effects on the data may result in effects on the MDS configuration derived from these data. In this section we discuss the hypothised effects of brand-decision involvement and product involvement on both aspects.

Two aspects of the dissimilarity data are considered to be sensitive to the involvement level of the consumer, namely the mean dissimilarity between the brands, and the variance of the dissimilarities. The effect of product involvement on the mean and the variance of the dissimilarity data is a priori unclear. A high level of brand-decision involvement is hypothised to result in a high mean of the dissimilarities, since a consumer who has a high brand-decision involvement level considers the brands to be relatively unequal. Little can be said in advance about the effect of brand-decision involvement on the variance of the dissimilarities.

The effects of involvement on the derived MDS configuration may concern aspects of the fit of the MDS model, the dimensionality of the perceptual map, and the estimated amount of error. Product involvement is hypothised to have a positive effect on the dimensionality (Gensch and Javalgi 1987). Consumers who are highly involved in the product are expected to use a large amount of product-relevant attributes when making the comparisons between brands. This will result in MDS solutions of high dimensionality. Furthermore, since their comparisons are well-considered, the fit of the model will increase, and the estimated standard error of the stimulus coordinates will be smaller. The effect of brand-decision involvement on the dimensionality of the perceptual map and the fit of the MDS model is expected to be positive (Gensch and Javalgi 1987), as a consumer who is highly involved with selecting a brand, will make clear distinctions between the brands. The influence of brand-decision involvement on the mean standard error is unclear.

4. DESIGN AND ANALYSIS OF THE STUDY

In order to investigate whether involvement influences the perceptions of consumers two empirical studies are performed with two different products, namely shampoos, expected to be a low-involvement product, and automobiles, expected to be a high-involvement product. The number of brands used are twelve in the case of automobiles and nine in the case of shampoos. For each product a sample of about 45 consumers has been drawn. The consumers are asked to make paired comparisons between all the pairs of brands (automobiles on a 9-point scale, and shampoos on a 7-point scale), and to answer the twenty-four items of the Mittal and Lee involvement scale (Appendix A).

At first, scales for product and brand-decision involvement are developed on the basis of the 24 items. Cronbach's Alpha is used to select the items and determine the internal reliability of the involvement scales. Secondly, respondents of both studies are classified into low, medium, and high involvement groups. This is done for product involvement as well as brand-decision involvement. Dissimilarity data will be aggregated across subjects within each of these involvement groups, and analysed with the probabilistic MDS model MULTISCALE (Ramsay 1991).

The theory of the analysis of dissimilarity data with MDS developed from metric MDS (Torgerson 1952), to non-metric MDS (Kruskal 1964; Shepard 1962), to analysis of individual differences (Carroll and Chang 1970), and into probabilistic MDS (Ramsay 1982). The last method has several important advantages over the deterministic MDS models. Probabilistic MDS takes the random properties of the dissimilarity data into account. In addition to stimulus coordinates, probabilistic MDS estimates variances of the stimulus points, and allows the researcher to test hypotheses. One of the main problems of deterministic models is that the selection of the dimensionality has to be made heuristically. Probabilistic models offer the opportunity to test whether an additional dimension improves the solution significantly. Since involvement may influence the dimensionality, the variances of the coordinates, as well as the coordinates themselves, we apply probabilistic MDS models in this study. We use the probabilistic model MULTISCALE, developed by Ramsay (Ramsay 1977, 1982 and 1991). This program computes maximum likelihood estimates of the point coordinates and standard errors of these estimates, which allows the researcher to obtain confidence regions of the stimulus coordinates (Ramsay 1978). Weinberg, Carroll, and Cohen (1984) showed that these estimates of standard errors provide a somewhat optimistic view of the actual statistical reliability of the solution, and consequently the confidence regions obtained from them are too small. On the other hand, Ramsay (1980) and Spence and Lewandowsky (1989) demonstrated the estimates of the point coordinates of MULTISCALE are fairly robust against small samples and outliers.

5. RESULTS

In order to obtain a reliable instrument for measuring product involvement and brand-decision involvement, the internal consistency of items of the Mittal and Lee questionnaire was investigated. We computed Cronbach's alpha for each of the scales, and deleted those items which reduced the reliability of the entire scale. The final product involvement scale for automobiles consists of eleven items (without item 6), and has a Cronbach's alpha of 0.88. The final brand-decision involvement for this product also contains eleven items (item 24 is omitted), and its internal reliability is 0.72. For shampoo the final product involvement scale contains eleven items (without item 6), alpha is 0.79, and for brand-decision involvement the final scale contains all twelve items with alpha equal to 0.82. We conclude that product involvement and brand-decision involvement have been measured reliably for both products. The scales are subsequently used for classification of the consumers into involvement groups. Scores for each consumer on the involvement scales are computed by adding the scores on the selected items. Then, three groups of about fifteen consumers are formed who have approximately the same level of product involvement. The same procedure is followed for brand-decision involvement (for both products).

TABLE 1

SCORES OF THE INVOLVEMENT GROUPS ON ASPECTS OF THE DISSIMILARITY DATA AND THE MDS SOLUTION

MULTISCALE allows the researcher to choose between three options for transforming the dissimilarity data, namely a scale, a power, and a spline transformation. Analysis of the data using a spline transformation does not yield the statistics of the null model. Therefore, we decided in advance not to use the spline transformation. Furthermore, it has to be decided whether the distribution of the dissimilarities is normal or lognormal, and it must be checked whether the data indicate that an individual differences MDS model is necessary. In advance, it is unclear which options have to be used in these empirical studies. Therefore, we performed eight analyses on the entire data set of automobiles and the entire data set of shampoos. In each case the dimensionality was set to two, but the data transformation option, the distribution option, and the individual differences option varied. For both data sets, the combination of options which yields the lowest score for the consistent version of the AIC (Akaike 1974), abbreviated to CAIC (Bozdogan 1987), will be applied in the following analyses of the involvement groups. The CAIC is a statistic which penalizes the log likelihood of the model more for the number of parameters estimated as compared to the AIC. In both cases a scale transformation, and a normal distribution, in combination with no individual weighting of the axes produces by far the lowest CAIC.

In the hypotheses section it is stated that product involvement and brand-decision involvement are hypothised to influence the dissimilarity data and the MDS solution. We distinguish between two features of the dissimilarity data, namely the mean, measured by averaging first within subjects and then between subjects, and the variance, measured by first computing the variance per subject and then averaging these variances. Furthermore, we consider the effects on the goodness of fit, the dimensionality, and the error of the perceptual map derived with probabilistic MDS. In order to be able to compare the goodness of fit values and the standard errors between the involvement groups, these measures are given for the two dimensional solutions. The goodness of fit is defined as twice the difference between the log likelihood of the two-dimensional solution and the log likelihood of the null model, in which each dissimilarity matrix is approximated by a constant. The higher the goodness of fit value, the more the two-dimensional model contributes to the explanation of the dissimilarity data as compared to the null model. In the case of probabilistic MDS the researcher has the opportunity to test which dimensionality fits the data best. That dimensionality is selected that yields the lowest CAIC (Bozdogan 1987). The CAIC tends to yield more parsimonious models than the AIC statistic. MULTISCALE computes not only coordinates of each stimulus, but in addition a standard error per stimulus per dimension. The mean of these standard errors can be interpreted as the over-all index of the uncertainty in the perceptual map.

Effects on the dissimilarity data

The results of the analyses of the effects of involvement on the dissimilarity data are presented in table 1. Though the mean of the dissimilarities does not differ significantly between the involvement groups, the effect of product involvement and brand-decision involvement on the mean appear to be positive. For each type of involvement and for each product the mean dissimilarity is lowest for the low involvement group. Especially in the case of automobiles the effect of product involvement is pronounced. Thus, consumers who are more involved with automobiles consider the brands to be less equal than less involved consumers do. Higher product involvement for automobiles also results in a decrease of the variance of the dissimilarities. This decrease in the variance is less evident for brand-decision involvement and for the shampoo data.

TABLE 2

INTERGROUP CORRELATIONS BETWEEN THE DISTANCE VECTORS DERIVED WITH MDS

Effects on the MDS solution

In the case of the automobiles study, the effects of product involvement and brand-decision involvement on the goodness of fit, the dimensionality and the mean standard error are partially contrary to the hypotheses (Table 1). Analysis of the high product involvement group results in a relatively low goodness of fit value, and low dimensionality. The low fit of this high-involvement group may be an indication of perceptual differences between highly involved consumers. The negative effect on the standard errors of the stimulus coordinates of product-involvement observed for the automobile data corresponds to the hypotheses. A rise in the brand-decision involvement level causes a rise in the dimensionality of the configuration. On the other hand, the effect of brand-decision involvement on the fit of the model, and the uncertainty of the stimulus coordinates is unclear. The medium involvement group has the lowest fit value, and the high involvement group has the highest mean standard error. In the case of shampoo the fit of the MDS model clearly increases if the level of product involvement or brand-decision involvement increases. The standard errors of both high involvement groups are significantly lower than those of the medium and low involvement groups. Conform Gensch and Javalgi (1987) and the basic features of the ELM (Petty and Cacioppo 1984 and 1986), we conclude that a high level of involvement causes a consumer to make well-considered evaluations of the shampoo brands on a relatively large number of characteristics, resulting in a perceptual map of relatively high dimensionality, higher goodness of fit level of the model, and low uncertainty of the estimated stimulus coordinates. The conclusions with respect to the automobile data are less unequivocal.

In order to investigate the effect of involvement on the perceptual map itself, we compared the two-dimensional solutions of the involvement groups. Since rotations of the maps do not affect the distances between the points (Ramsay 1982), the perceptual maps can be rotated to maximum agreement. To accomplish this, we apply the module FMATCH of the statistical package PC-MDS, which is based on an algorithm developed by Cliff (1966), for procrustes analysis. The correlations between the distance vectors of the perceptual maps are presented in Table 2. In general the correlations are high, which means the perceptual maps do not differ strongly between the involvement groups. However, as expected, in each case the lowest intergroup correlation is between the two extreme involvement groups.

6. CONCLUSION AND DISCUSSION

In general, we conclude that our data support the hypothesis that involvement influences brand dissimilarity perceptions. However, their are no clear differences between the effects of product involvement and those of brand-decision involvement on both the dissimilarity ratings and the MDS solutions. Higher involvement appears to increase the mean dissimilarity ratings. The effect of involvement on the variance of the dissimilarities is unclear; for automobiles it is negative, for shampoo no influence is observed. We also found some empirical evidence that an increase in involvement level results in a higher goodness of fit of the two-dimensional MDS solution for the low-involvement product shampoo only. In addition to that, involvement appears to have a positive effect on the dimensionality of the perceptual maps recovered, and a negative effect on the uncertainty of the stimulus coordinates. The final perceptual maps do not strongly differ between the involvement groups. In the framework of the ELM, these findings support the hypothesis that a highly involved consumer is more motivated and able to process information centrally, resulting in a well-considered perception based on a relatively large amount of product-relevant attributes. The differences found in our study between involvement groups were modest, however, and the effects of involvement on consumer perceptions of brands and their attributes may have been overrated in the literature.

Research on a broader range of consumer aspects that may have an impact on the perceptual maps is needed, and the effects of involvement have to be investigated for a larger set of products and brands, and with larger samples. In addition to involvement, aspects like risk perception and brand familiarity may play an important role. It may even be advisable to develop and apply models which incorporate these aspects (Chatterjee and DeSarbo 1992).

APPENDIX A

ITEMS OF THE MITTAL AND LEE INVOLVEMENT MODEL

7. REFERENCES

Akaike, Hirotugu (1974), "A new look at the statistical model identification," IEEE Transactions on Automatic Control, AC-19, 716-723.

Bloch, Peter H., and Marsha L. Richins (1983), "A theoretical model for the study of product importance perceptions," Journal of Marketing, 47 (Summer), 69-81.

Bozdogan, Hamparsum (1987), "Model selection and Akaike's information criterion (AIC): the general theory and its analytical extensions," Psychometrika, 52, 345-370.

Carroll, J. Douglas, and J.J. Chang (1970), "Analysis of individual differences in Multidimensional Scaling via an N-way generalization of "Eckart-Young" decomposition," Psychometrika, 35, 283-319.

Celsi, Richard L., and Jerry C. Olson (1988), "The role of involvement in attention and comprehension processes," Journal of Consumer Research, 15 (September), 210-224.

Chatterjee, Rabikar, and Wayne S. DeSarbo (1992), "Accommodating the effects of brand unfamiliarity in the Multidimensional Scaling of preference data," Marketing Letters, 3, 85-99.

Cliff, Norman (1966), Orthogonal rotation to congruence, Psychometrika, 31, 33-42.

Cooper, Lee G. (1983), "A review of Multidimensional Scaling in Marketing Research," Applied Psychological Measurement, 7, 427-450.

Engel, James F., and Roger D. Blackwell (1982), Consumer Behavior. New York: Dryden Press.

Gensch, Dennis H. and Rajshekhar G. Javalgi (1987), "The influence of involvement on disaggregate attribute choice models," Journal of Consumer Research, 14 (June), 71-82.

Green, Paul E. (1975), "Marketing applications of MDS: assessment and outlook," Journal of Marketing, 39, 24-31.

Humphreys, Marie A. (1982), "Data collection effects on nonmetric Multidimensional Scaling solutions," Educational and Psychological Measurement, 42, 1005-1022.

Jain, Arun K., and Christian Pinson (1976), "The effect of order of presentation of similarity judgments on Multidimensional Scaling results: an empirical examination," Journal of Marketing Research, 13 (November), 435-439.

Johnson, Michael D., Donald R. Lehmann, and Daniel R. Horne (1990), "The effects of fatigue on judgements of interproduct similarity," International Journal of Research in Marketing, 7, 35-43 and 53-56.

Kapferer, Jean-Nodl, and Gilles Laurent (1985), "Consumers' involvement profile: new empirical results," In: Annual conference proceedings of the Association for Consumer Research, Hirschman and Holbrook, editors, 290-295.

Kapferer, Jean-Nodl, and Gilles Laurent (1986), "Consumer involvement profiles: a new practical approach to consumer involvement," Journal of Advertising Research, 25 (January), 48-56.

Kirmani, Amna (1990), "The effect of perceived advertising costs on brand perceptions," Journal of Consumer Research, 17 (September), 160-171.

Kruskal, Joseph B. (1964), "Multidimensional Scaling by optimizing goodness of fit to a nonmetric hypothesis," Psychometrika, 29, 1-27 and 115-129.

Laurent, Gilles, and Jean-Nodl Kapferer (1985), "Measuring consumer involvement profiles," Journal of Marketing Research, 22, 41-53.

MacKenzie, Scott B., and Richard A. Spreng (1992), "How does motivation moderate the impact of central and peripheral processing on brand attitudes and intentions?" Journal of Consumer Research, 18, 519-529.

Malhotra, Naresh K. (1990), "Comment on: The effects of fatigue on judgements of interproduct similarity by M.D. Johnson, D.R. Lehmann, and D.R. Horne," International Journal of Research in Marketing, 7, 45-51.

Malhotra, Naresh K., Arun K. Jain, and Christian Pinson (1988), "The robustness of MDS configurations in the case of incomplete data," Journal of Marketing Research, 25 (February), 95-102.

Miniard, Paul W., Sunil Bhatla, Kenneth R. Lord, Peter R. Dickson, and H. Rao Unnava (1991), "Picture-based persuasion processes and the moderating role of involvement," Journal of Consumer Research, 18 (June), 92-107.

Mittal, Banwari (1989), "Measuring purchase-decision involvement," Psychology & Marketing, 6 (Summer), 147-162.

Mittal, Banwari, and Myung-Soo Lee (1989), "A causal model of consumer involvement," Journal of Economic Psychology, 10, 363-389.

Petty, Richard E., and John T. Cacioppo (1984), "The effects of involvement on responses to argument quantity and quality: central and peripheral routes to persuasion," Journal of Personality and Social Psychology, 46, 69-81.

Petty, Richard E., and John T. Cacioppo (1986), Communication and persuasion: central and peripheral routes to attitude change. New York: Springer.

Petty, Richard E., John T. Cacioppo, and David Schumann (1983), "Central and peripheral routes to advertising effectiveness: the moderating role of involvement," Journal of Consumer Research, 10 (September), 135-146.

Ramsay, James O. (1977), "Maximum likelihood estimation in Multidimensional Scaling," Psychometrika, 42, 241-266.

Ramsay, James O. (1978), "Confidence regions for Multidimensional Scaling analysis," Psychometrika, 43, 145-160.

Ramsay, James O. (1980), "Some small sample results for maximum likelihood estimation in Multidimensional Scaling," Psychometrika, 45, 139-146.

Ramsay, James O. (1982), "Some statistical approaches to Multidimensional Scaling data," Journal of the Royal Statistical Society, 145 (A), 285-312.

Ramsay, James O. (1991), MULTISCALE manual.Ritchie, J.R. Brent (1974), "An exploratory analysis of the nature and extent of individual differences in perceptions," Journal of Marketing Research, 11, 41-49.

Shepard, Roger N. (1962), "The analysis of proximities: Multidimensional Scaling with an unknown distance function," Psychometrika, 27, 125-140 and 219-246.

Spence, Ian (1972), "A Monte Carlo evaluation of three nonmetric Multidimensional Scaling algorithms," Psychometrika, 37, 461-486.

Spence, Ian, and Stephan Lewandowsky (1989), "Robust Multidimensional Scaling," Psychometrika, 54, 501-513.

Torgerson, Warren S. (1952), "Multidimensional Scaling: I. Theory and method," Psychometrika, 17, 401-419.

Weeks, David G., and Peter M. Bentler (1979), "A comparison of linear and monotone Multidimensional Scaling models," Psychological Bulletin, 86, 349-354.

Weinberg, Sharon L., J. Douglas Carroll, and Harvey S. Cohen (1984), "Confidence regions for INDSCAL using the jackknife and bootstrap techniques," Psychometrika, 49, 475-491.

Whipple, Thomas W. (1976), "Variation among Multidimensional Scaling solutions: an examination of the effect of data collection differences," Journal of Marketing Research, 13 (February), 98-103.

Zaichkowsky, Judith L. (1985), "Measuring the involvement construct," Journal of Consumer Research, 12 (December), 341-352.

----------------------------------------