A Comparison of Some Methods of Cluster Analysis with SPSS - Dissertation Example

duction to Classification and Clustering Statistical analysis is the process by which those conducting research and analysing data, can determine who or what within a dataset, fit certain patterns and trends. There is always a dependent or prominent variable which is affected by independent variables under different analytical circumstances and then there is clustering a group of people, for example, who may have similar buying propensities or who respond the same way to a certain dosage in a medical treatment (Norusis 361).

As Burns and Burns describe it, cluster analysis classifies ‘a mountain of information into manageable meaningful piles’ (552). Clustering into groups helps in identifying and classifying particular categories into a membership, from which a classification rule is determined. In a simple description of cluster analysis, it is a generic name for mathematical operations which determine what classified objects fit closely in a group (Romesburg 2). Analysis conducted on a batch of rocks as the main group, will show through analysis that some are classified as simple round pebbles, others are quartz, rough diamonds (hopefully) or fool’s gold (typical luck).

Characteristics of the rocks then reclassify into smaller clustered groups, depending on the goal of the research (2). Linkage between the variables, the cases and the clusters are a main proponent of cluster analysis (Burns and Burns 1). Classification analysis is used more often in regular research analysis than people realize and there are several ways of approaching classifications, as reviewed in the next section. Information and marketing research has found ways to conduct all types of cluster sampling, for example, in order to learn more about what is happening in their market with consumers, their purchasing habits, and where these are occurring.

One popular form of research is through area sampling, where clusters are done by geographic designations such as north, northwest, south, southwest, and so on, or by metropolitan statistical areas (MSAs), such as cities, streets, and regional divisions (Hair, Bush, and Ortinau 352). Whatever the sampling is, cluster sample provides that sampling clustered units are divided into exclusive groupings where each cluster is considered a representative of mutually similar components (Zikmund 708).

A more common term used in the marketing research field is segmentation when referring to a population group of customers and this can also be cluster sampled by customers in different cities to find out which cities are alike in consumer purchasing (Churchill and Iacobucci 820). In psychology, clustering is a process of putting together groups of people, based on their responses to variables, rather than grouping those variables, such as found in factor analysis (Field 1). From that point, Euclidean distance determines the geometric distance between two objects, also known as cases.

In the cases where there are some negative and some positive differences, the distances are squared, therefore providing a positive distance. This is because a negative distance, squared, becomes a positive. A positive distance, squared, remains a positive distance. At the end of squaring all the distances, then they are all summed up and then the square