Analyzing relationships between covariates and trajectories coded as a sequence of states

What is it about?

This paper propose a methodological framework for analyzing the relationship between state sequences and covariates. Based on the pairwise dissimilarities, it looks at how the covariates explain the discrepancy of the sequences, which makes it possible to develop a series of statistical significance–based analysis tools.

Featured Image

Why is it important?

The proposed methodology opens new perspectives besides the traditional cluster-based approach. In short, this cluster-based approach consists of associating each trajectory in a given set to some related ideal type. From a descriptive standpoint, this approach has proven to be effective in uncovering the underlying structure of a set of sequences, which makes the data easier to understand. However, relying on clusters for studying the relationship between sequences and their context can be criticized on the basis that reducing the set of sequences to a limited number of standard trajectories is a rather crude approximation and would lead to considering deviations from the standard inside a cluster as non explained error terms. As a result of this approximation, wrong conclusions may be drawn about relationships between the sequences and their context. On the other hand, the approach we propose here takes into account explicitly how the individual characteristics affect the trajectory followed by each individual.