Profile Analysis

In psychological research, there are many instances in which the goal is to classify people as belonging to certain groups. Personality researchers often search for evidence of personality types, some intelligence researchers search for evidence that people exhibit different types of intelligences, and so on. The term profile analysis is used to represent a variety of techniques that share the common goal of empirically classifying individual observations into distinguishable groups based on common characteristics measured by observed variables. Strictly speaking, it is not a statistical technique because no inferences are being made to population parameters. Rather, profile analysis is a data reduction technique. It is conceptually similar to factor analysis, but the difference is that the focus in profile analysis is on grouping people (or observations) on the basis of common traits measured by observed variables.

The two major approaches to profile analysis are quite different from one another conceptually, but they share the same goal of grouping cases based on observed variables. The first approach is based on correlation methods, such as Pearson's r or Spearman's rho. The goal under this approach is to group cases together that show similar patterns of spikes and dips across variables (i.e., the same shape), regardless of the absolute level of the scores (see Figure 1). Under the correlational approach, Persons A and B would be classified as belonging to one group (or profile) and Persons C and D to another.

By contrast, a second approach to profiling participants is based on measures of distance, such as Euclidean distance or Mahalanobis's distance (see Figure 1). Here, the emphasis is on creating groups based on the extent to which case scores are close in absolute value (i.e., level), regardless of the similarity of pattern shape. Under the distance approach, Persons B and C would be classified as belonging to one group and Persons A and D another.

The Case Study and the Data

Consider a practical example in which students were administered an experimental Advanced Placement test in Psychology. The test consisted of items primarily tapping one of four cognitive processing skills: memory processing, analytical processing, creative processing, or practical processing. The researchers were interested in examining the extent to which students showed different profiles of strengths and weaknesses across these four processing skills. Consequently, a Q-factor analysis (a correlation-based profile analysis) was performed on the students who took the test to determine the extent to which different profiles of achievement were observed. Using a principal component analysis (of the cases, not the variables), the researchers arrived at a solution yielding

Figure 1 Examples of Different Score Profiles

[Page 789]

six general profiles in the data. Overall, the researchers were able to group each of the 1,262 cases in the data sets as belonging to one of the six empirically distinguishable profiles.

...