Quantitative Research (Cluster and Regression Analysis) Essay

Quantitative Research (Cluster and Regression Analysis) - Essay Example In most of the marketing data sets like the one taken now for analysis none of these two conditions hold. Data were collected from the database provided by the chain store, the loyalty program of which were researched upon. In any supermarket store loyalty, the location of the store plays a major role as a determinant for the loyalty of the customer (Messinger & Narasimhan, 1997; Engel et al., 1995). In this study the location of the three stores chosen would definitely have affected the purchasing behavior of the customers. Hence the study becomes weak in its identification of the data source. Another weakness of the data has been observed in the varying proportion of the different categories of buying which will have an influence on the application of the cluster analysis technique. The study has not identified the total number of customers of the three stores and hence it would be difficult to comment on the randomness of the sample selected. A comparison of the total number of customers and the number constituting the sample size would have thrown some light on the comparability of the loyal customers between the customers who shopped generally during the period under study and the number of customers who opted to use the loyalty program. Basis for Collection of Information - Clustering Variables The collection of information and clustering considering the percentage of total share of wallet within product categories instead of taking into account the total purchases would have been a much better presentation of data under the research method of clustering analysis. The variables selected are far too general to form an opinion on the customer loyalty. The clustering lacks seriously because of the massiveness of data considered under the general clustering variables. For sure these clustering variables would have been subjected to behavioral benchmarking. It would be interesting to recall the behavioral factors like shopping frequency, tolerance of price increase etc. (Lacey, 2003) Validation of the Clusters The study has used the numerical taxonomy process to group the members into segments (Bunn, 1993; McKelvey, 1975; Punj and Stewart, 1983) However no clarity appears to be in sight in determining the range of potential market structures. Initially the number of groups ranged from two to eleven. Although the study has used the appropriate testing and analysis methods like scree testing, discriminant analysis and regression analysis to arrive at a particular number of groups as cluster groups, there is the lack of a scientific variation among the different groups evolved for study. There are possibilities that a slight change in the scaling would have vitiated the results especially in the middle range groups. This may be either due to problems of scaling as observed by Long (1997) or due to large volume of data analysed. According to Long (1997) "Scaling is a common cause of problems when numerical or interdependence methods are utilized, with the ratio between the largest standard deviation and the smallest standard deviation considered heuristically predictive of the likelihood and size of