In determining the correct multivariate statistical analysis to use, executive management must first understand the meaning of multivariate statistics. Multivariate statistics is the area of statistics that deals with observations made on many variables. The primary objective is to study how the variables are related to one another and how they work in combination to distinguish between the cases in which they made observations. Companies require market research firms to use multivariate statistical techniques, such as Factor Analysis, Multidimensional Scaling, and Cluster Analysis, to find connections between their service or products (one variable) to their target customer bases (another variable) or any other variable. These multivariate statistical techniques are approaches to directly understanding the relationship between all the variables to maximize the company's growth, revenues, and profits. In 1985, the United States Drug Administration (USDA) established a study of women’s nutrition using a multivariate statistical technique to explain the quantitative nutritional habits of American women. One of the questions the USDA wanted to be answered was if the women in the study could be partitioned or classified as similar individuals because there are possibilities of developing group-specific educational protocols, and one educational approach will not contain all women. The statistical approach they used was the Cluster Analysis method due to its ability to describe similar groups within a large sample. Understanding the different multivariate statistical techniques and choosing the right one helps a company’s executive management decide to enter into developing new products or any project that benefits the company’s success.
Factor Analysis
Factor Analysis is another multivariate statistical technique businesses use to reduce many variables to a lesser number of factors by extracting the maximum standard variances from all variables and then putting them into a standard score. Factor analysis is also a part of the general linear model that includes the assumptions of (1) including relevant variables into the analysis, (2) validating a direct relationship, (3) having no multi-collinearity, and (4) establishing if a precise correlation between variables and factors exists. There are different methods, under the Factor Analysis method, utilized by researchers to extract elements from the data sets. The most commonly used extraction method by researchers is the Principal Component Analysis method which obtains the maximum variance, places them into the first factor, then removes the explained variance by the first factors. After removing the explained variance, this method starts extracting the maximum variation from the second factor to the last factor. The second most commonly used Factor Analysis method by researchers is the Common Factor Analysis process which extracts the common variance, puts them into elements, and includes a unique variety of all the variables. Researchers use the Factor Analysis method for investigating variable relationships for complex concepts such as socioeconomic status, dietary patterns, or psychological scales and concepts that are not easily measured directly by collapsing many variables into a few interpretable underlying factors to determine the similarity in patterns of responses associated with a latent variable. The Factor Analysis method has assumptions, but it also has objectives too. Researchers have defined the Factor Analysis Method objectives as figuring out how many factors are necessary to discuss the variable sets, finding the associative value of each variable with the set of common factors, explaining the common factors, and determining the importance of each factor in every observation. One goal of the Factor Analysis method is to provide a valuable discussion of the variance for the set of variables inputted into the analysis by factors (hidden original dimensions). Verizon Wireless, the telecom giant, utilized the Factor Analysis method to determine what the consumers want in their products, such as their cell phones and the type of plan that will benefit them and their families. They used this method to assist their researchers in pinpointing the percentage and perception of the consumers by conducting surveys or marketing a sample or demo of their products. The company performed factor analysis for a year on the sales of a black cell phone, then added a blue, silver, and red cell phone, tracking the factors for each sale and comparing it to the sales of the black cell phone to determine the ideal product that most consumers will purchase. Verizon Wireless uses this method to conclude the relationship between the variables they need to continue the different styles and colors of cell phones. Companies can learn from Verizon Wireless using the Factor Analysis method for their studies and possibly use the Factor Analysis method as a starting point for the data they are seeking but transition over to the Cluster Analysis process to complete their studies.
Multidimensional Scaling Method (MDS)
The Multidimensional Scaling Method (MDS), from a non-technical point of view, provides a visual representation of the pattern of proximity (i.e., similarities or distances) among a set of objects. MDS is useful information when considering the pricing and branding of products, and a company will need this information either for present or future studies. Additionally, MDS can help businesses determine the market share of their products compared to similar products of competitors. MDS has a close relationship with the Cluster Analysis method. MDS is also a way that captures the similarity of objects, but MDS fails to identify similarities in the relationship because there is no significant variance for the MDS to determine and obtain the data. An excellent feature of MDS is the ability to analyze any similarity or distance algorithm. MDS is very familiar with marketing research because MDS can determine the different angles of underlying perceptions of different products or brands. This piece of information is exquisite to know and add to a company’s knowledge database. Overall, MDS is right for determining similarities if you are looking for similarities.
Cluster Analysis Method
The Cluster Analysis method originated in 1932 through the field of anthropology by Driver and Kroeber, but in 1938 and 1939, Zubin and Robert Tyron brought this approach into the psychology field. In 1943, Cattell made the Cluster Analysis Method famous for trait theory classification by studying personality psychology. The Cluster Analysis method has many uses for a company to enter, or expand, into an industry. In marketing, a company can use the Cluster Analysis method to identify and reach out to potential markets. Regarding products and development, the Cluster Analysis method can help create new goods and improve them based on consumer preferences. The Cluster Analysis method, or clustering, is an excellent tool for any business that can properly use its elements. It is very helpful for marketing or designing products centered on different customer needs to complete the task of exploratory data mining necessary for a company. The Cluster Analysis method works by taking a source of data and breaking it down into different clusters (groups). These clusters serve to categorize the data by whatever criteria the company chooses, such as demographics. This method is suitable for business because the Cluster Analysis process is not bound by one algorithm to determine what constitutes a cluster and how efficient the study can become in finding the groups. Companies can use it to provide services or products that are more in tune with the needs of particular consumers. If seeking knowledge discovery, a company should use the Cluster Analysis method to gain new markets or strengthen their market share in current markets. The Cluster Analysis process will take a statistical approach in grouping similar kinds of variables into respective categories but can modify the pre-processing of the data and cluster parameters until the results reach the desired properties intended for the study. It identifies an individual group that’s forming and whether the group is strong or weak as the analysis repeats over time. WiseWindow used the Cluster Analysis method to improve companies’ social media content. WiseWindow is part of KPMG and designed a patented web logical system that masters cluster analysis. WiseWindow helped one of their music industry clients analyze over 25 million comments on the top 500 musical acts in the U.S. This Cluster Analysis method displayed the rise and fall of intensity to 84 aspects of different artists, such as their music or stage presence. The Cluster Analysis process found a strong correlation between record sales and social media activity. The Cluster Analysis method splits up data clusters that are meaningful and useful to the company. If a company seeks to look for those target clusters, then this method should have a proper structure of their desired data. A company can use clustering as a launch point for data summation of possible target groups, pattern recognition, information retrieval, and data mining. The Cluster Analysis method just does what is natural to human beings; separating things into groups (clustering) and assigning things into groups (classification).
The Cluster Analysis method is the better technique for companies to use in determining the different products and if entering an industry is suitable for the enterprise too. With all the information researched, the Cluster Analysis method seems like the technique that gives companies better results of executive leadership’s expectations. Now factor analysis is a technique a company can start, but not a technology to depend on the better results as the Cluster Analysis method.