Widgecorp III There a Number of Techniques

WidgeCorp III

There a number of techniques for performing multivariate analysis. These include factor analysis, multidimensional scaling and cluster analysis. The first of these, factor analysis, is used to determine "something about the nature of the independent variables that affect the dependent variables," without actually measuring the independent variables (Darlington, n.d.).

DeCoster (1998) notes that there are two types of factor analysis: exploratory factor analysis and confirmatory factor analysis. As DeCoster points out "Factor analyses are performed by examining the pattern of correlations (or covariances) between the observed measures. Measures that are highly correlated are likely influenced by the same factors."

An example in the real world is outlined by Goldberg (n.d.). In this example, an energy company wanted explanations for trends in customer energy usage. The study featured sixteen different factors ranging from bill services to cost/rates issues to type of house. This example is not very good, because the conclusion is just that these factors are affecting the energy usage and that was the starting point of the study. However, in examples that are better explained we would have the researcher outline specifically "This factor is a strong influencer because it has an impact on these other factors." An example would be the number of people in the house, which might impact on the type of house, the presence of modern technology in the home, the importance of service issues and other variables. The objective is to determine the underlying variables that have the most impact. These will be the ones with the highest correlation with other variables.

If there is a variable that has no correlation with the other ones, then that is a factor with low impact on the dependent variable.

The second type of multivariate analysis is multi-dimensional scaling. MDS is a "set of data analysis techniques that display the structure of distance-like data as a geometrical picture" (Young, n.d.). There are different types of multidimensional scaling. These include classical MDS, metric MDS, nonmetric CDMS, replicated MDS and weighted MDS.

One type of MDS that is commonly used in business is found in marketing departments, and is known as the perceptual map. This technique maps out physically the perceptions that customers have of different products, based on where those perceptions fit on a matrix. The perceptual map for snack food, for example, would be derived from survey data about how the customers perceive various different snacks, including our own. We could use a basic matrix like price-quality, or any other pair of variables. The map would then show us where our snack food fits in comparison to those of our competitors.

This technique is used when a visual representation is valuable -- for example when we are comparing our product vs. A large number of competitors. The most sophisticated MDS techniques can allow us to test perceptions of pairs of snack food items or pairs of attributes, with the result being that we can determine what the most important variables are for consumers of snack foods by identifying the variables that have the highest correlations -- like if a certain product routinely outscores other products, we can drill down to the different variables and determine why that product is so successful. This adds value in both product development and marketing -- we already know what is successful but knowing why helps us to gain competitive advantage in the areas that are most important to consumers.

The final multivariate analysis technique is cluster analysis. This is similar to the other techniques but is seeks to identify clusters. So for example, if we were to look at the fast food market, and take the top 20 or 50 firms, we could use clustering to identify the most successful variables by examining the interdependence of each variable. We would probably find that "burgers" is something where there is a cluster. But we might find that Sonic, being a drive-in, is an outlier, while emphasis on drive-through service was a cluster. We could also find that there is a correlation between perceived low price and perceived value, and that many of the most successful firms are clustered where those two variables intersect.

Again, the objective of cluster analysis is to find the clusters where two variables meet that have the highest rates of success. In snack food, we have basic figures on what types of snacks are most popular, but we can use cluster analysis to determine what attribute clusters are in the market, and what products/companies feature those attributes. We could determine that a price/value cluster exists, a salt/fat cluster exists or whatever pair, but clustering can help to us determine where those clusters are and who is exploiting them.