Functional Clustering for the Social Science Toolkit

• In Event: Recent Advances in Mathematical Sociology

Abstract

Social science research often seeks to compare trends over time or space, such as life-course transitions or developmental trajectories. However, standard approaches such as sequence analysis or group-based trajectory modeling are limited in their ability to classify trends in continuous data. In this paper, we introduce functional clustering, a method widely used in biostatistics and epidemiology but rarely applied in the social sciences. Functional clustering can be used to identify types of trajectories based on their overall shape and rate of change. We describe the steps of the method, including the advantages and disadvantages of various methodological choices. We demonstrate the use of functional clustering by analyzing racial residential segregation as a function of distance to identify distinct macro- and micro-segregation typologies. By extending this method to the social sciences, we provide a flexible and robust framework for analyzing dynamic social processes that unfold across time and space.

Authors

• Elizabeth Roberto, Rice University

• Caroline Wolski, Rice University