Examine correlations between lexical habits and other variables. Often in lexical analysis you will find that certain characteristics of a person's speech correlates with other traits of that person. After collecting a group of people's lexical data, use statistics to analyze the frequency of certain words of interest in that person's speech. Then see if these people who frequently use words of interest have any external, non-lexical traits in common. For example, people who tend to use positive words significantly more often than negative words tend to be happier.
Classify conversations according to the statistical lexical properties. Especially for conversations between two people, it is often true that the speech of each person engaging in the conversation converges. Using statistics, you can determine words of interest and examine how the people in the conversation match on these words. Often using statistical hypothesis testing is appropriate when testing if two people differ in their speech patterns. For instance, couples who frequently use words like "we" and "us" tend to be more stable than couples who do not.
Categorize a person's overall speech pattern. Some experts in lexical analysis have begun to use discriminant analysis, a form of statistical classification, to organize speeches into types. Discriminant analysis is a technique that can take speech data in as input and output a function that clearly separates different speeches. One example of this is asking individuals to explain some concept to the best of their ability. After receiving the explanation, you score the responses on an appropriate scale. Input the lexical data as independent data and the scores as dependent data. After running discriminant analysis, you will have criteria for what kind of speech gives high scores or low scores.