Methods of Classification - Preface to Special Section

Methods and Concepts of Classification Classification comes in many forms and disguises. Cars are classified in trucks and dune buggies, humans are classified in the good ones, the bad ones, and the ugly ones, voters try to create groups of political candidates such as acceptable ones and no-way candidates, where, before knowing the candidates, it is unknown which groups exist and which candidate goes into which group. Medical diagnoses are made after performing classification processes. Decisions about which item to order from a restaurant menu follow classification procedures. When perceiving new objects, classification takes place to reduce information overload. The same applies in attempts to memorize as many as possible from a list of phone numbers or nonsense syllables. From a person perspective (Bergman & Magnusson, 1997; von Eye & Bergman, 2003), methods of classification are of great importance. The person perspective proposes that groups or even individuals exist that differ from other groups in two respects. First, they differ in such parameters as means or standard deviations. This proposition reflects what is known as the differential perspective discussed, for instance, by Stern (1903). Second, and this is unique to the person perspective, even if they don't differ in means or standard deviations, they differ in functioning. That is, covariance structures, predictive structures, or process characteristics may differ across groups. Methods of classification either use existing groups or help identify such groups if they are unknown before data collection. Existing groups can be examined in regard to possible differential functioning. Groups can be identified with differential functioning in mind, or they can be identified using standard procedures and then examined in regard to functioning. Current methods of classification can be classified into three groups. The first involves assigning individual cases to pre-existing groups. The question is only which group does a thus far not classified individual belong to. The groups themselves are considered given. They will not be changed based on the number and characteristics of those who are assigned to them. For example, if five students are classified as suffering from Attention Deficit Disorder (ADD), the nosological category ADD will remain unchanged. The second group of classification methods involves distinguishing between groups or classes in the space of other variables than the ones used to define the groupings. If groups differ in the space of these other variables, they are often considered valid (e.g., Tubman et al., 1990). Methods of analysis used in this domain include discriminant function analysis and logistic regression. The third group of classification methods involves creating groupings from existing data. Here, methods of cluster analysis and latent class analysis are most prominent. The Special Section on Methods of Classification In this Special Section on Methods of Classification, we intend to show that there is more to classification than assigning individual cases to a priori-defined existing groupings or classes, discriminating cases from different groupings, or creating a priori unknown groupings or classes from existing data. This Special section presents a selection of the aspects under which classification can be discussed and performed. The second and the third of the three classical perspectives listed above will be discussed, but also new perspectives. This Special Section includes eight articles. We do not propose a grouping for these articles because, fortunately, they cover quite different aspects of classification. The order in which the articles are presented is thus rather arbitrary. The first article in this Special section was contributed by Baillargeon, Tremblay (both at the Universite de Montreal, Canada), Willms (University of New Bunswick, Canada), Romano, Lee, and Wu (all at the Universite de Montreal, Canada). …